intro.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Welcome to TapeAgents!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "**TapeAgents** is a framework that leverages a structured, replayable log (**Tape**) of the agent session to facilitate all stages of the LLM Agent development lifecycle. In TapeAgents, the agent reasons by processing the tape and the LLM output to produce new thoughts, actions, control flow steps and append them to the tape. The environment then reacts to the agent’s actions by likewise appending observation steps to the tape.\n",
    "\n",
    "In this tutorial, you will learn:\n",
    "- how to create TapeAgents using the low-level API\n",
    "- run and resume TapeAgents\n",
    "- have one TapeAgent reuse another TapeAgent's tape as training data\n",
    "\n",
    "In upcoming versions of this tutorial, you will also learn: \n",
    "- how to make a team TapeAgent with subagents\n",
    "- how to build TapeAgents using available high-level APIs\n",
    "- how to build a TapeAgent that streams partial steps\n",
    "\n",
    "Other tutorials and examples will cover:\n",
    "- code execution and browser use\n",
    "- finetuning\n",
    "- the TapeAgents apps (Studio and Browser)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Setup\n",
    "We're assuming that you already installed the project through the `make setup` or the jupyter notebook is running in the context of the project. If not, please refer to the [README](README.md) for more detailed instructions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:23.372986Z",
     "iopub.status.busy": "2025-02-24T13:10:23.372480Z",
     "iopub.status.idle": "2025-02-24T13:10:23.385457Z",
     "shell.execute_reply": "2025-02-24T13:10:23.384604Z"
    }
   },
   "outputs": [],
   "source": [
    "# Now set the OPENAI_API_KEY environment variable to your API key.\n",
    "\n",
    "import os\n",
    "\n",
    "if \"OPENAI_API_KEY\" not in os.environ:\n",
    "    os.environ[\"OPENAI_API_KEY\"] = \"<your-api-key>\"\n",
    "    # os.environ[\"OPENAI_ORGANIZATION\"] = \"\" # optional\n",
    "today = \"2024-09-17\"  # fixed date for reproducible tests\n",
    "\n",
    "\n",
    "# If you prefer to skip the OpenAI setup and not make any LLM calls, you can use ones from the cache.\n",
    "# it will work instead of the real LLM fine as long as the prompts are not changed.\n",
    "# Uncomment the following lines to use the cache:\n",
    "#\n",
    "# from tapeagents import llms\n",
    "# import os\n",
    "# llm_cache_path = \"tests/res/intro_notebook/tapedata.sqlite\"\n",
    "# if not os.path.exists(llm_cache_path):\n",
    "#     llm_cache_path = f\"../{llm_cache_path}\"\n",
    "# assert os.path.exists(llm_cache_path)\n",
    "# llms._REPLAY_SQLITE = llm_cache_path"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Your first TapeAgent"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this section, we will build the simplest possible \"hello world\" agent. We will then go through all the new concepts that you need to know to understand the code. This section is quite long, but with the solid foundation you acquire here other TapeAgent tutorials will be easy to process.\n",
    "\n",
    "Without further ado, here's the code!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:23.390548Z",
     "iopub.status.busy": "2025-02-24T13:10:23.390186Z",
     "iopub.status.idle": "2025-02-24T13:10:28.060765Z",
     "shell.execute_reply": "2025-02-24T13:10:28.060061Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/oleh.shliazhko/TapeAgents/.venv/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final tape: {\n",
      "  \"metadata\": {\n",
      "    \"id\": \"d0d10c89-65b7-4a3c-993b-abc63e61d9b7\",\n",
      "    \"parent_id\": \"8ebf62cb-64ed-46d6-b407-df9c80283be9\",\n",
      "    \"author\": \"Agent\",\n",
      "    \"author_tape_id\": null,\n",
      "    \"n_added_steps\": 2,\n",
      "    \"error\": null,\n",
      "    \"result\": {}\n",
      "  },\n",
      "  \"context\": null,\n",
      "  \"steps\": [\n",
      "    {\n",
      "      \"metadata\": {\n",
      "        \"id\": \"421b7b7c-cc61-45e7-8b77-b1fc894dfe2c\",\n",
      "        \"prompt_id\": \"\",\n",
      "        \"node\": \"\",\n",
      "        \"agent\": \"\",\n",
      "        \"llm\": \"\",\n",
      "        \"other\": {}\n",
      "      },\n",
      "      \"kind\": \"user\",\n",
      "      \"content\": \"Tell me about Vulcan in 3 sentences\"\n",
      "    },\n",
      "    {\n",
      "      \"metadata\": {\n",
      "        \"id\": \"dab90836-249a-4e5b-b0bf-d565de8529c0\",\n",
      "        \"prompt_id\": \"e85f2f34-57eb-4514-a842-c2f12d673726\",\n",
      "        \"node\": \"main\",\n",
      "        \"agent\": \"Agent\",\n",
      "        \"llm\": \"default\",\n",
      "        \"other\": {}\n",
      "      },\n",
      "      \"kind\": \"assistant\",\n",
      "      \"content\": \"Vulcan is a fictional planet in the \\\"Star Trek\\\" universe, known as the home of the Vulcan species, including the iconic character Spock. The Vulcans are characterized by their logical minds, emotional control, and advanced technology, having a rich cultural history that emphasizes peace and science. The planet itself features a harsh landscape with mountainous regions and arid environments, and it plays a crucial role in the wider narrative of the \\\"Star Trek\\\" saga.\"\n",
      "    },\n",
      "    {\n",
      "      \"metadata\": {\n",
      "        \"id\": \"1e580704-3106-49f1-b762-10b747a89c4a\",\n",
      "        \"prompt_id\": \"e85f2f34-57eb-4514-a842-c2f12d673726\",\n",
      "        \"node\": \"main\",\n",
      "        \"agent\": \"Agent\",\n",
      "        \"llm\": \"default\",\n",
      "        \"other\": {}\n",
      "      },\n",
      "      \"kind\": \"set_next_node\",\n",
      "      \"next_node\": \"main\"\n",
      "    }\n",
      "  ]\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "from tapeagents.agent import Agent, Node\n",
    "from tapeagents.core import Prompt, SetNextNode\n",
    "from tapeagents.dialog_tape import AssistantStep, DialogTape, UserStep\n",
    "from tapeagents.llms import LiteLLM, LLMStream\n",
    "from tapeagents.prompting import tape_to_messages\n",
    "\n",
    "llm = LiteLLM(model_name=\"gpt-4o-mini\")\n",
    "\n",
    "\n",
    "class MainNode(Node):\n",
    "    name: str = \"main\"\n",
    "\n",
    "    def make_prompt(self, agent: Agent, tape: DialogTape) -> Prompt:\n",
    "        # Render the whole tape into the prompt, each step is converted to message\n",
    "        return Prompt(messages=tape_to_messages(tape))\n",
    "\n",
    "    def generate_steps(self, agent: Agent, tape: DialogTape, llm_stream: LLMStream):\n",
    "        yield AssistantStep(content=llm_stream.get_text())  # Generate new step from the LLM output stream.\n",
    "        yield SetNextNode(next_node=\"main\")  # Which node to execute next, more on that later\n",
    "\n",
    "\n",
    "agent = Agent[DialogTape].create(llm, nodes=[MainNode()])\n",
    "start_tape = DialogTape(steps=[UserStep(content=\"Tell me about Vulcan in 3 sentences\")])\n",
    "final_tape = agent.run(start_tape).get_final_tape()  # agent will start executing the first node\n",
    "print(f\"Final tape: {final_tape.model_dump_json(indent=2)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now let's learn about tapes, steps, prompts, llm streams, nodes and agents."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tape\n",
    "\n",
    "The fundamental concept of the TapeAgents is the `Tape`, a comprehensive semantic level log of the agent's session. A `Tape` contains a context and a sequence of `Step` objects. As you can see, a TapeAgent runs by adding steps (such as `UserStep` or `AssistantStep`) to the _tape_. This example uses the `DialogTape` tape, which is a basic tape for user-assistant conversations. Let's see what are the possible steps in a `DialogTape`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:28.063008Z",
     "iopub.status.busy": "2025-02-24T13:10:28.062749Z",
     "iopub.status.idle": "2025-02-24T13:10:28.078690Z",
     "shell.execute_reply": "2025-02-24T13:10:28.078394Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tapeagents.core.Tape[Union[DialogContext, NoneType], Union[UserStep, ToolResult, SystemStep, AssistantThought, SetNextNode, Pass, Call, Respond, FinalStep, AssistantStep, ToolCalls]]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# We use Python generics to instantiate many different Tape types by\n",
    "# specifying different Context and Step types. In the output of this cell,\n",
    "# look at Union[UserStep, AssistantStep, ...]\n",
    "# for the list of possible step types in the DialogTape.\n",
    "DialogTape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some of these steps should be familiar to you. `UserStep`, `AssistantStep`, `SystemStep` and `ToolResult` correspond to `role=user`, `role=assistant`, `role=system` and `role=tool` LLM API messages respectively. `ToolCalls` and `AssistantThought` correspond to assistant messages where the LLM requests a tool call or produces an intermediate thought that is not meant to be shown to the user. `SetNextNode` and `Pass` are TapeAgent's internal step to control which node it should run at the next iteration (more on this below)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Prompt format; LLMs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We use the industry-standard \"chat.completions\" prompt format in TapeAgents: a list of user/assistant/system/tool messages plus tool schemas."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:28.080106Z",
     "iopub.status.busy": "2025-02-24T13:10:28.079993Z",
     "iopub.status.idle": "2025-02-24T13:10:28.082494Z",
     "shell.execute_reply": "2025-02-24T13:10:28.082044Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'id': FieldInfo(annotation=str, required=False, default_factory=<lambda>),\n",
       " 'tools': FieldInfo(annotation=Union[list[dict], NoneType], required=False, default=None),\n",
       " 'messages': FieldInfo(annotation=list[dict], required=False, default_factory=list),\n",
       " 'token_ids': FieldInfo(annotation=list[int], required=False, default_factory=list)}"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Almost all classes in TapeAgents are Pydantic base models.\n",
    "# This allows easy validation, serialization and instrospection. For example,\n",
    "# here we are able to list all the fields in the Prompt model.\n",
    "Prompt.model_fields"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The LLMs in TapeAgent take `Prompt` and return an `LLMStream` object. The `LLMStream` object can be used both to fast-forward to the complete response text and to stream partial outputs step by step."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:28.084376Z",
     "iopub.status.busy": "2025-02-24T13:10:28.084246Z",
     "iopub.status.idle": "2025-02-24T13:10:33.727263Z",
     "shell.execute_reply": "2025-02-24T13:10:33.726750Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Certainly! Here’s a simple example of a \""
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hello, World!\" program in Java:\n",
      "\n",
      "```java\n",
      "public class Hello"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "World {\n",
      "    public static void main(String[] args) {\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "        System.out.println(\"Hello, World!\");\n",
      "    }\n",
      "}\n",
      "```\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "To run this program:\n",
      "\n",
      "1. Save the code in a file named `Hello"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "World.java`.\n",
      "2. Open your command line or terminal.\n",
      "3. Navigate to"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " the directory where the file is saved.\n",
      "4."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Compile the code using the command: `javac HelloWorld.java`\n",
      "5."
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Run the compiled program using the command: `java HelloWorld`\n",
      "\n",
      "You should see"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " the output:\n",
      "\n",
      "```\n",
      "Hello, World!\n",
      "```None\n",
      "------------------------------\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Certainly! Here is a simple \"Hello, World!\" program written in C:\n",
      "\n",
      "```c\n",
      "#include <stdio.h>\n",
      "\n",
      "int main() {\n",
      "    printf(\"Hello, World!\\n\");\n",
      "    return 0;\n",
      "}\n",
      "```\n",
      "\n",
      "### Explanation:\n",
      "- `#include <stdio.h>`: This line includes the standard input-output library, which is necessary for using the `printf` function.\n",
      "- `int main()`: This defines the main function, which is the entry point of the program.\n",
      "- `printf(\"Hello, World!\\n\");`: This line prints \"Hello, World!\" followed by a newline character to the standard output.\n",
      "- `return 0;`: This indicates that the program has finished executing successfully.\n",
      "\n",
      "To compile and run this code:\n",
      "1. Save it in a file named `hello.c`.\n",
      "2. Open a terminal and navigate to the directory where you saved the file.\n",
      "3. Compile the program using a C compiler, like `gcc`:\n",
      "   ```\n",
      "   gcc hello.c -o hello\n",
      "   ```\n",
      "4. Run the compiled program:\n",
      "   ```\n",
      "   ./hello\n",
      "   ```\n",
      "\n",
      "You should see `Hello, World!` printed on the screen.\n"
     ]
    }
   ],
   "source": [
    "llm_stream = LiteLLM(model_name=\"gpt-4o-mini-2024-07-18\", stream=True)\n",
    "\n",
    "# Streaming\n",
    "prompt = Prompt(messages=[{\"role\": \"user\", \"content\": \"Write hello world in Java\"}])\n",
    "for event in llm_stream.generate(prompt):\n",
    "    print(event.chunk, end=\"\")\n",
    "\n",
    "# No streaming\n",
    "# (note: you can not use Prompt object for more than 1 LLM call in TapeAgents)\n",
    "prompt = Prompt(messages=[{\"role\": \"user\", \"content\": \"Write hello world in C\"}])\n",
    "print(\"\\n\" + \"-\" * 30)\n",
    "print(llm_stream.generate(prompt).get_text())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In the example above we use the easiest way to create a prompt from the tapes: `tape_to_messages`. Under the hood, this method uses `step.llm_dict()` method of all non-control steps in the tape to create the prompt:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:33.729294Z",
     "iopub.status.busy": "2025-02-24T13:10:33.729092Z",
     "iopub.status.idle": "2025-02-24T13:10:33.732954Z",
     "shell.execute_reply": "2025-02-24T13:10:33.732527Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'kind': 'user', 'content': 'hi AI!'}\n",
      "{'kind': 'assistant', 'content': 'hello human'}\n",
      "[{'role': 'user', 'content': 'hi AI!'}, {'role': 'assistant', 'content': 'hello human'}]\n"
     ]
    }
   ],
   "source": [
    "print((user := UserStep(content=\"hi AI!\")).llm_dict())\n",
    "print((assistant := AssistantStep(content=\"hello human\")).llm_dict())\n",
    "print(tape_to_messages(DialogTape(steps=[user, assistant])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A key priority in TapeAgents is making use of the data that running the agent generates. To make this possible, some TapeAgent LLMs know how to make their finetuning data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:33.734687Z",
     "iopub.status.busy": "2025-02-24T13:10:33.734557Z",
     "iopub.status.idle": "2025-02-24T13:10:34.490369Z",
     "shell.execute_reply": "2025-02-24T13:10:34.490026Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--- ALL TEXT ---\n",
      "<|user|>\n",
      "Say bla 3 times and foo 2 times<|end|>\n",
      "<|assistant|>\n",
      "Sure! Let me say bla bla bla foo foo<|end|>\n",
      "<|endoftext|>\n",
      "--- PREDICTED CHARACTERS ---\n",
      "Sure! Let me say bla bla bla foo foo<|end|>\n",
      "<|endoftext|>\n"
     ]
    }
   ],
   "source": [
    "from tapeagents.llms import LLMOutput, TrainableLLM\n",
    "\n",
    "trainable_llm = TrainableLLM(\n",
    "    base_url=\"\",  # we only use the tokenizer from the model here, no need for a base_url for inference\n",
    "    model_name=\"microsoft/Phi-3.5-MoE-instruct\",\n",
    "    tokenizer_name=\"microsoft/Phi-3.5-MoE-instruct\",\n",
    ")\n",
    "\n",
    "simple_tape = DialogTape(\n",
    "    steps=[\n",
    "        UserStep(content=\"Say bla 3 times and foo 2 times\"),\n",
    "        AssistantStep(content=\"Sure! Let me say bla bla bla foo foo\"),\n",
    "    ]\n",
    ")\n",
    "\n",
    "prompt = Prompt(messages=tape_to_messages(simple_tape[:1]))  # type: ignore\n",
    "output = agent.make_llm_output(simple_tape, index=1)\n",
    "text = trainable_llm.make_training_text(prompt=prompt, output=output)\n",
    "print(\"--- ALL TEXT ---\")\n",
    "print(text.text)\n",
    "print(\"--- PREDICTED CHARACTERS ---\")\n",
    "print(text.output_text)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Node"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A node represents an uninterruptible atom of TapeAgent's computation. When TapeAgents runs a node, it uses its two main functions: `make_prompt` to create an LLM Prompt from the tape and `generate_steps` to create new steps from the LLM output. To build a node, you can subclass `Node` and override these functions. Note that `generate_steps` must be a generator, a design choice we made to make TapeAgents a streaming-friendly framework. \n",
    "\n",
    "Let's see what the node from the above example can do."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:34.495274Z",
     "iopub.status.busy": "2025-02-24T13:10:34.494783Z",
     "iopub.status.idle": "2025-02-24T13:10:35.641197Z",
     "shell.execute_reply": "2025-02-24T13:10:35.640081Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Raw node prompt:\n",
      "id='147e928f-3914-4b00-8b98-fb0ac5cd4beb' tools=None messages=[{'role': 'user', 'content': 'Hi, AI!'}] token_ids=[]\n",
      "\n",
      "Steps produced by node's generator:\n",
      "[AssistantStep(metadata=StepMetadata(id='01187e00-3328-4efe-8daf-afe9516d7fb4', prompt_id='', node='', agent='', llm='', other={}), kind='assistant', content='Hello, human!'), SetNextNode(metadata=StepMetadata(id='9b5f7a2e-1425-4a42-87e5-8109b0a7281d', prompt_id='', node='', agent='', llm='', other={}), kind='set_next_node', next_node='main')]\n",
      "\n",
      "Produced Steps:\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"0066c409-6267-42f7-abf9-8ac5dfd1fca4\",\n",
      "    \"prompt_id\": \"\",\n",
      "    \"node\": \"\",\n",
      "    \"agent\": \"\",\n",
      "    \"llm\": \"\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"assistant\",\n",
      "  \"content\": \"Hello! How can I assist you today?\"\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"9fe33868-2dde-4178-bda6-c0a70b91973d\",\n",
      "    \"prompt_id\": \"\",\n",
      "    \"node\": \"\",\n",
      "    \"agent\": \"\",\n",
      "    \"llm\": \"\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"set_next_node\",\n",
      "  \"next_node\": \"main\"\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "from tapeagents.llms import LLMEvent\n",
    "\n",
    "\n",
    "class MainNode(Node):\n",
    "    name: str = \"main\"\n",
    "\n",
    "    def make_prompt(self, agent, tape: DialogTape) -> Prompt:\n",
    "        return Prompt(messages=tape_to_messages(tape))\n",
    "\n",
    "    def generate_steps(self, agent, tape, llm_stream: LLMStream):\n",
    "        yield AssistantStep(content=llm_stream.get_text())\n",
    "        yield SetNextNode(next_node=\"main\")  # Continue to the same node\n",
    "\n",
    "\n",
    "node = MainNode()\n",
    "\n",
    "# Let's run \"make_prompt\" in isolation.\n",
    "prompt = node.make_prompt(agent=None, tape=DialogTape(steps=[UserStep(content=\"Hi, AI!\")]))\n",
    "print(f\"Raw node prompt:\\n{prompt}\\n\")\n",
    "\n",
    "\n",
    "# Now, let's run \"generate_steps\" in isolation.\n",
    "# We need to construct a fake LLMStream to do that.\n",
    "def _generator():\n",
    "    yield LLMEvent(output=LLMOutput(content=\"Hello, human!\"))\n",
    "\n",
    "\n",
    "stream = LLMStream(_generator(), Prompt())\n",
    "step_stream = node.generate_steps(agent=None, tape=DialogTape(), llm_stream=stream)\n",
    "print(f\"Steps produced by node's generator:\\n{list(step_stream)}\\n\")\n",
    "\n",
    "# When the agent runs a node, it is equivalent to the following three steps:\n",
    "# Step 1: make a prompt\n",
    "start_tape = DialogTape(steps=[UserStep(content=\"Hi, AI!\")])\n",
    "prompt = node.make_prompt(agent, start_tape)\n",
    "# Step 2: construct the LLMStream from the prompt (happens inside the agent)\n",
    "stream = llm.generate(prompt)\n",
    "# Step 3: generate steps that the agent will then add to the tape\n",
    "print(\"Produced Steps:\")\n",
    "for step in node.generate_steps(agent, start_tape, stream):\n",
    "    print(step.model_dump_json(indent=2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Agent and its nodes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The TapeAgent agent iteratively runs the nodes and appends the steps generated by each node to the tape. To select which next node to run, internally a TapeAgent computes the **tape view** object. The Tape remains the only **state** that the agent uses, the view only represents its content in a way that is convenient for the agent to use.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:35.645621Z",
     "iopub.status.busy": "2025-02-24T13:10:35.645163Z",
     "iopub.status.idle": "2025-02-24T13:10:35.654557Z",
     "shell.execute_reply": "2025-02-24T13:10:35.653939Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1: \n",
      "2: main\n",
      "3: act\n"
     ]
    }
   ],
   "source": [
    "from tapeagents.core import StepMetadata\n",
    "from tapeagents.view import TapeViewStack\n",
    "\n",
    "# The \"top\" view in the tape view stack is the view of the current agent.\n",
    "# Initially `top.last_node` is empty and the agent will run the first node from its list\".\n",
    "tape1 = DialogTape(steps=[UserStep(content=\"Hi, AI!\")])\n",
    "last_node = TapeViewStack.compute(tape1).top.last_node\n",
    "print(f\"1: {last_node}\")\n",
    "assert last_node == \"\"\n",
    "\n",
    "\n",
    "# When the agent computes the view, it updates `top.last_node` with the node from the latest agent step\n",
    "# The agent will search the next node after the last_node in its nodes list.\n",
    "tape2 = DialogTape(\n",
    "    steps=[\n",
    "        UserStep(content=\"Hi, AI!\"),\n",
    "        AssistantStep(metadata=StepMetadata(prompt_id=\"123\", node=\"main\"), content=\"AI here, how I can help?\"),\n",
    "    ]\n",
    ")\n",
    "last_node = TapeViewStack.compute(tape2).top.last_node\n",
    "print(f\"2: {last_node}\")\n",
    "assert last_node == \"main\"\n",
    "\n",
    "# The SetNextNode step on the tape changes `top.next_node` to the value of the `next_node` field in the SetNextNode step.\n",
    "# The agent will use this value\n",
    "tape3 = DialogTape(\n",
    "    steps=[\n",
    "        UserStep(content=\"Hi, AI!\"),\n",
    "        AssistantStep(metadata=StepMetadata(prompt_id=\"123\"), content=\"AI here, how I can help?\"),\n",
    "        SetNextNode(next_node=\"act\"),\n",
    "    ]\n",
    ")\n",
    "next_node = TapeViewStack.compute(tape3).top.next_node\n",
    "print(f\"3: {next_node}\")\n",
    "assert next_node == \"act\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By default the agent stops after the last node has produced an `Action` step. The action steps are the steps by which the agent requests information from the environment. For example, `AssistantStep` is an `Action` as it indicates the agent awaits the user response, `ToolCalls` is an action requesting tool call results. Let's look at all possible steps in `DialogTape` tape and see which of them are actions, observations and thoughts."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:35.658395Z",
     "iopub.status.busy": "2025-02-24T13:10:35.657843Z",
     "iopub.status.idle": "2025-02-24T13:10:35.662693Z",
     "shell.execute_reply": "2025-02-24T13:10:35.662134Z"
    }
   },
   "outputs": [],
   "source": [
    "from tapeagents.core import Action, Observation, Pass, Thought\n",
    "from tapeagents.dialog_tape import AssistantThought, ToolCalls, ToolResult\n",
    "\n",
    "assert all([issubclass(step_class, Action) for step_class in [AssistantStep, ToolCalls]])\n",
    "assert all([issubclass(step_class, Thought) for step_class in [AssistantThought, SetNextNode, Pass]])\n",
    "assert all([issubclass(step_class, Observation) for step_class in [UserStep, ToolResult]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we are ready to look at a simplified summary of the corner-stone `agent.run` algorithm.\n",
    "\n",
    "1. Compute the new tape view\n",
    "2. Choose the active agent (more on multi-agent TapeAgents later)\n",
    "3. Choose the active node\n",
    "4. Run the node and add steps on the tape\n",
    "5. If the last node yielded an action, then stop, else repeat.\n",
    "\n",
    "`agent.run` returns an `AgentStream` object which allows iterating through the agent's steps (or partial steps when streaming) and fast-forwardin to the complete new tape with `get_final_tape`."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Converse with a TapeAgent"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's continue the conversation with the agent that previously responded to us with the Vulcan definition from the StarTrek.\n",
    "Remember, the session is stored in the tape **final_tape**."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:35.664977Z",
     "iopub.status.busy": "2025-02-24T13:10:35.664679Z",
     "iopub.status.idle": "2025-02-24T13:10:37.420060Z",
     "shell.execute_reply": "2025-02-24T13:10:37.419157Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"84369d82-1c65-4c15-9634-4a5827245070\",\n",
      "    \"parent_id\": \"b89698d0-d29e-454c-9f7b-b89e829ac016\",\n",
      "    \"author\": \"Agent\",\n",
      "    \"author_tape_id\": null,\n",
      "    \"n_added_steps\": 2,\n",
      "    \"error\": null,\n",
      "    \"result\": {}\n",
      "  },\n",
      "  \"context\": null,\n",
      "  \"steps\": [\n",
      "    {\n",
      "      \"metadata\": {\n",
      "        \"id\": \"421b7b7c-cc61-45e7-8b77-b1fc894dfe2c\",\n",
      "        \"prompt_id\": \"\",\n",
      "        \"node\": \"\",\n",
      "        \"agent\": \"\",\n",
      "        \"llm\": \"\",\n",
      "        \"other\": {}\n",
      "      },\n",
      "      \"kind\": \"user\",\n",
      "      \"content\": \"Tell me about Vulcan in 3 sentences\"\n",
      "    },\n",
      "    {\n",
      "      \"metadata\": {\n",
      "        \"id\": \"dab90836-249a-4e5b-b0bf-d565de8529c0\",\n",
      "        \"prompt_id\": \"e85f2f34-57eb-4514-a842-c2f12d673726\",\n",
      "        \"node\": \"main\",\n",
      "        \"agent\": \"Agent\",\n",
      "        \"llm\": \"default\",\n",
      "        \"other\": {}\n",
      "      },\n",
      "      \"kind\": \"assistant\",\n",
      "      \"content\": \"Vulcan is a fictional planet in the \\\"Star Trek\\\" universe, known as the home of the Vulcan species, including the iconic character Spock. The Vulcans are characterized by their logical minds, emotional control, and advanced technology, having a rich cultural history that emphasizes peace and science. The planet itself features a harsh landscape with mountainous regions and arid environments, and it plays a crucial role in the wider narrative of the \\\"Star Trek\\\" saga.\"\n",
      "    },\n",
      "    {\n",
      "      \"metadata\": {\n",
      "        \"id\": \"1e580704-3106-49f1-b762-10b747a89c4a\",\n",
      "        \"prompt_id\": \"e85f2f34-57eb-4514-a842-c2f12d673726\",\n",
      "        \"node\": \"main\",\n",
      "        \"agent\": \"Agent\",\n",
      "        \"llm\": \"default\",\n",
      "        \"other\": {}\n",
      "      },\n",
      "      \"kind\": \"set_next_node\",\n",
      "      \"next_node\": \"main\"\n",
      "    },\n",
      "    {\n",
      "      \"metadata\": {\n",
      "        \"id\": \"ff8db3ad-6baa-49a4-a0d5-59e9a23ed01c\",\n",
      "        \"prompt_id\": \"\",\n",
      "        \"node\": \"\",\n",
      "        \"agent\": \"\",\n",
      "        \"llm\": \"\",\n",
      "        \"other\": {}\n",
      "      },\n",
      "      \"kind\": \"user\",\n",
      "      \"content\": \"No, I mean Vulcan the company\"\n",
      "    },\n",
      "    {\n",
      "      \"metadata\": {\n",
      "        \"id\": \"4a9f184a-96df-4c37-aa9f-77d95a62be2e\",\n",
      "        \"prompt_id\": \"f44f6196-1ef5-4bc0-88c3-ae77cdd8ec17\",\n",
      "        \"node\": \"main\",\n",
      "        \"agent\": \"Agent\",\n",
      "        \"llm\": \"default\",\n",
      "        \"other\": {}\n",
      "      },\n",
      "      \"kind\": \"assistant\",\n",
      "      \"content\": \"Vulcan Inc. is a private investment firm founded by the late Paul Allen, co-founder of Microsoft, in 1986. The company focuses on various sectors, including technology, real estate, and philanthropy, with an emphasis on innovative solutions to global challenges and environmental sustainability. Vulcan has also been involved in notable projects such as the Museum of Pop Culture in Seattle and various initiatives aimed at conservation and scientific research.\"\n",
      "    },\n",
      "    {\n",
      "      \"metadata\": {\n",
      "        \"id\": \"c9d45cae-0fa1-44f3-85e9-35637af12ce9\",\n",
      "        \"prompt_id\": \"f44f6196-1ef5-4bc0-88c3-ae77cdd8ec17\",\n",
      "        \"node\": \"main\",\n",
      "        \"agent\": \"Agent\",\n",
      "        \"llm\": \"default\",\n",
      "        \"other\": {}\n",
      "      },\n",
      "      \"kind\": \"set_next_node\",\n",
      "      \"next_node\": \"main\"\n",
      "    }\n",
      "  ]\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "tape_to_continue = final_tape + [UserStep(content=\"No, I mean Vulcan the company\")]\n",
    "continued_tape = agent.run(tape_to_continue).get_final_tape()\n",
    "print(continued_tape.model_dump_json(indent=2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Note that the agent is able to continue talking to you thanks for `SetNextNode(next_node=\"main\")` step that `generate_steps` produced. If you try to remove this step as an exercise, the agent will crash because there is only one node."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Tape rendering\n",
    "\n",
    "LLM agents create a lot of data that can be overwhelming to process. In TapeAgents we render the tape with the associated prompts and outputs into a more readable HTML for you. To make this work, we store prompts and outputs in an SQLite database every time you call `agent.run()`.\n",
    "\n",
    "Here's how to use tape rendering in the notebook:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:37.423427Z",
     "iopub.status.busy": "2025-02-24T13:10:37.423063Z",
     "iopub.status.idle": "2025-02-24T13:10:37.620923Z",
     "shell.execute_reply": "2025-02-24T13:10:37.620655Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>.basic-renderer-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; white-space: pre-wrap; color: black;}.basic-prompt-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; color: black;}.episode-row { display: flex; align-items: end; }.agent-column { width: 50%; }.user-column { width: 25%; }.annotator-column { width: 25%; }.inner-tape-container { display: flex }.inner-tape-indent { width: 10%; }.inner-tape { width: 90%; }.agent_node { text-align: center; position: relative; margin: 12px 0; }.agent_node hr { border: 1px solid #e1e1e1; margin: 0 !important;}.agent_node span { position: absolute; right: 0; top: -0.6em; background: white; color: grey !important; padding: 0 10px; }</style><style>.observation { background-color: #baffc9;; }.error_observation { background-color: #dd0000; }.action { background-color: #cccccc; }.thought { background-color: #ffffdb; }.call { background-color: #ffffff; }.respond { background-color: #ffffff; }.step-header { margin: 2pt 2pt 2pt 0 !important; }.step-text { font-size: 12px; white-space: pre-wrap; word-wrap: break-word;}</style><h3 style='margin: 2px'>Metadata</h3><details> <summary>id: 84369d82-1c65-4c15-9634-4a5827245070</summary> <div class='basic-renderer-box'>author: Agent\n",
       "author_tape_id: null\n",
       "error: null\n",
       "id: 84369d82-1c65-4c15-9634-4a5827245070\n",
       "n_added_steps: 2\n",
       "parent_id: b89698d0-d29e-454c-9f7b-b89e829ac016\n",
       "result: {}\n",
       "</div> </details><h3 style='margin: 2px'>Steps</h3><div class='basic-renderer-box observation'><h4 class='step-header'>[0] User</h4><pre class='step-text'>kind: user\n",
       "\n",
       "Tell me about Vulcan in 3 sentences</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 16 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>user: Tell me about Vulcan in 3 sentences</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 98 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='Vulcan is a fictional planet in the \"Star Trek\" universe, known as the home of the Vulcan species, including the iconic character Spock. The Vulcans are characterized by their logical minds, emotional control, and advanced technology, having a rich cultural history that emphasizes peace and science. The planet itself features a harsh landscape with mountainous regions and arid environments, and it plays a crucial role in the wider narrative of the \"Star Trek\" saga.', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[1] Assistant</h4><pre class='step-text'>kind: assistant\n",
       "\n",
       "Vulcan is a fictional planet in the \"Star Trek\" universe, known as the home of the Vulcan species, including the iconic character Spock. The Vulcans are characterized by their logical minds, emotional control, and advanced technology, having a rich cultural history that emphasizes peace and science. The planet itself features a harsh landscape with mountainous regions and arid environments, and it plays a crucial role in the wider narrative of the \"Star Trek\" saga.</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[2] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: main\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[3] User</h4><pre class='step-text'>kind: user\n",
       "\n",
       "No, I mean Vulcan the company</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 126 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: Vulcan is a fictional planet in the \"Star Trek\" universe, known as the home of the Vulcan species, including the iconic character Spock. The Vulcans are characterized by their logical minds, emotional control, and advanced technology, having a rich cultural history that emphasizes peace and science. The planet itself features a harsh landscape with mountainous regions and arid environments, and it plays a crucial role in the wider narrative of the \"Star Trek\" saga.\n",
       "--\n",
       "user: No, I mean Vulcan the company</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 89 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='Vulcan Inc. is a private investment firm founded by the late Paul Allen, co-founder of Microsoft, in 1986. The company focuses on various sectors, including technology, real estate, and philanthropy, with an emphasis on innovative solutions to global challenges and environmental sustainability. Vulcan has also been involved in notable projects such as the Museum of Pop Culture in Seattle and various initiatives aimed at conservation and scientific research.', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[4] Assistant</h4><pre class='step-text'>kind: assistant\n",
       "\n",
       "Vulcan Inc. is a private investment firm founded by the late Paul Allen, co-founder of Microsoft, in 1986. The company focuses on various sectors, including technology, real estate, and philanthropy, with an emphasis on innovative solutions to global challenges and environmental sustainability. Vulcan has also been involved in notable projects such as the Museum of Pop Culture in Seattle and various initiatives aimed at conservation and scientific research.</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[5] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: main\n",
       "</pre></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.display import HTML\n",
    "\n",
    "from tapeagents.renderers import render_tape_with_prompts\n",
    "from tapeagents.renderers.pretty import PrettyRenderer\n",
    "\n",
    "HTML(render_tape_with_prompts(continued_tape, PrettyRenderer()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Your TapeAgent with planning and tools"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's build a TapeAgent that plans and acts. We will be using OpenAI function calling capabilities in this example."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:37.622187Z",
     "iopub.status.busy": "2025-02-24T13:10:37.622102Z",
     "iopub.status.idle": "2025-02-24T13:10:47.732275Z",
     "shell.execute_reply": "2025-02-24T13:10:47.731104Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Run the agent!\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"b1e0b05b-b7ae-4c77-b58e-bb930fc88188\",\n",
      "    \"prompt_id\": \"b6c39afc-09e9-42b0-9b43-fe33338119b8\",\n",
      "    \"node\": \"plan\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"assistant_thought\",\n",
      "  \"content\": \"1. Identify that the user is asking about a company named \\\"Vulcan\\\" and wants information about it.\\n\\n2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\\n\\n3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\\n\\n4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.\"\n",
      "}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"3def80ea-863c-4646-91f7-fa4a6a010fc1\",\n",
      "    \"prompt_id\": \"c2f8e9d9-42c9-4d6f-982e-7926a14c4e31\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"assistant\",\n",
      "  \"tool_calls\": [\n",
      "    {\n",
      "      \"function\": {\n",
      "        \"name\": \"get_stock_ticker\",\n",
      "        \"arguments\": \"{\\\"company_name\\\":\\\"Vulcan\\\"}\"\n",
      "      },\n",
      "      \"id\": \"call_al69pbS0aE2LrEXiYTarQ2ZJ\",\n",
      "      \"type\": \"function\"\n",
      "    }\n",
      "  ]\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"9db137e6-de74-4945-a00c-6bc963f095fb\",\n",
      "    \"prompt_id\": \"c2f8e9d9-42c9-4d6f-982e-7926a14c4e31\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"set_next_node\",\n",
      "  \"next_node\": \"act\"\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"ffbbebb0-54ff-4a9a-9ff0-516a82f784ee\",\n",
      "    \"prompt_id\": \"\",\n",
      "    \"node\": \"\",\n",
      "    \"agent\": \"\",\n",
      "    \"llm\": \"\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"tool\",\n",
      "  \"content\": \"VMC\",\n",
      "  \"tool_call_id\": \"call_al69pbS0aE2LrEXiYTarQ2ZJ\"\n",
      "}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"43ab478b-6fd3-460a-9221-f7a7d001e773\",\n",
      "    \"prompt_id\": \"9d1e797f-3bab-4e36-8347-31e2328af9ea\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"assistant\",\n",
      "  \"tool_calls\": [\n",
      "    {\n",
      "      \"function\": {\n",
      "        \"name\": \"get_stock_data\",\n",
      "        \"arguments\": \"{\\\"symbol\\\":\\\"VMC\\\",\\\"start_date\\\":\\\"2024-09-01\\\",\\\"end_date\\\":\\\"2024-09-16\\\"}\"\n",
      "      },\n",
      "      \"id\": \"call_s0VrNLHmJksS9Mgvh4tgmXCb\",\n",
      "      \"type\": \"function\"\n",
      "    }\n",
      "  ]\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"9f5140b9-a114-409b-8222-81fc2dfcad4f\",\n",
      "    \"prompt_id\": \"9d1e797f-3bab-4e36-8347-31e2328af9ea\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"set_next_node\",\n",
      "  \"next_node\": \"act\"\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"9f1e2822-9744-4e2b-91d2-d64279bc8a97\",\n",
      "    \"prompt_id\": \"\",\n",
      "    \"node\": \"\",\n",
      "    \"agent\": \"\",\n",
      "    \"llm\": \"\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"tool\",\n",
      "  \"content\": [\n",
      "    [\n",
      "      \"2024-09-03\",\n",
      "      239.02000427246094\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-04\",\n",
      "      234.8300018310547\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-05\",\n",
      "      232.16000366210938\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-06\",\n",
      "      231.97000122070312\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-09\",\n",
      "      231.8300018310547\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-10\",\n",
      "      232.9499969482422\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-11\",\n",
      "      232.9499969482422\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-12\",\n",
      "      236.27000427246094\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-13\",\n",
      "      237.47000122070312\n",
      "    ]\n",
      "  ],\n",
      "  \"tool_call_id\": \"call_s0VrNLHmJksS9Mgvh4tgmXCb\"\n",
      "}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"f06ad1c3-9989-45a8-a7b1-313e36567325\",\n",
      "    \"prompt_id\": \"93da0b34-e3f7-4293-9e8b-6a3b9f910e7a\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"assistant\",\n",
      "  \"content\": \"Vulcan Materials Company, trading under the stock ticker symbol VMC, is a major producer of construction aggregates and a leading producer of other construction materials. Recently, its stock prices have shown some fluctuations, with prices ranging from approximately $231.83 to $239.02 between September 1 and September 13, 2024. This reflects a stable yet slightly volatile market performance, indicative of the construction industry's dynamics and economic conditions.\"\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"fe547291-104c-47a6-a08a-3d3b2ae8a105\",\n",
      "    \"prompt_id\": \"93da0b34-e3f7-4293-9e8b-6a3b9f910e7a\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"set_next_node\",\n",
      "  \"next_node\": \"plan\"\n",
      "}\n",
      "Final tape:\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<style>.basic-renderer-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; white-space: pre-wrap; color: black;}.basic-prompt-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; color: black;}.episode-row { display: flex; align-items: end; }.agent-column { width: 50%; }.user-column { width: 25%; }.annotator-column { width: 25%; }.inner-tape-container { display: flex }.inner-tape-indent { width: 10%; }.inner-tape { width: 90%; }.agent_node { text-align: center; position: relative; margin: 12px 0; }.agent_node hr { border: 1px solid #e1e1e1; margin: 0 !important;}.agent_node span { position: absolute; right: 0; top: -0.6em; background: white; color: grey !important; padding: 0 10px; }</style><style>.observation { background-color: #baffc9;; }.error_observation { background-color: #dd0000; }.action { background-color: #cccccc; }.thought { background-color: #ffffdb; }.call { background-color: #ffffff; }.respond { background-color: #ffffff; }.step-header { margin: 2pt 2pt 2pt 0 !important; }.step-text { font-size: 12px; white-space: pre-wrap; word-wrap: break-word;}</style><h3 style='margin: 2px'>Metadata</h3><details> <summary>id: 2e73b731-ef41-4faa-8045-3db2b0f5e6e2</summary> <div class='basic-renderer-box'>author: Agent\n",
       "author_tape_id: null\n",
       "error: null\n",
       "id: 2e73b731-ef41-4faa-8045-3db2b0f5e6e2\n",
       "n_added_steps: 2\n",
       "parent_id: 0487e8dc-28c3-4f22-851d-e530977d3c9d\n",
       "result: {}\n",
       "</div> </details><h3 style='margin: 2px'>Steps</h3><div class='basic-renderer-box observation'><h4 class='step-header'>[0] User</h4><pre class='step-text'>kind: user\n",
       "\n",
       "Tell me about Vulcan in 3 sentences</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 98 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_ticker\",\n",
       "      \"description\": \"Get company stock ticker from its name.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"company_name\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"company_name\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_data\",\n",
       "      \"description\": \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"symbol\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"start_date\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"end_date\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"symbol\",\n",
       "          \"start_date\",\n",
       "          \"end_date\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "user: Write a natural language plan on how to use tools help the user. Output a list of numbered items, like 1., 2., 3., etc.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 122 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\\n\\n2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\\n\\n3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\\n\\n4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box thought'><h4 class='step-header'>[1] Thought: AssistantThought</h4><pre class='step-text'>kind: assistant_thought\n",
       "\n",
       "1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\n",
       "\n",
       "2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\n",
       "\n",
       "3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\n",
       "\n",
       "4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 205 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_ticker\",\n",
       "      \"description\": \"Get company stock ticker from its name.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"company_name\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"company_name\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_data\",\n",
       "      \"description\": \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"symbol\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"start_date\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"end_date\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"symbol\",\n",
       "          \"start_date\",\n",
       "          \"end_date\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\n",
       "\n",
       "2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\n",
       "\n",
       "3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\n",
       "\n",
       "4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.\n",
       "--\n",
       "user: Follow the plan you created to earlier. When you are done, respond to the user.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 0 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=None, role='assistant', tool_calls=[ChatCompletionMessageToolCall(function=Function(arguments='{\"company_name\":\"Vulcan\"}', name='get_stock_ticker'), id='call_al69pbS0aE2LrEXiYTarQ2ZJ', type='function')], function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[2] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_ticker\n",
       "    arguments: '{\"company_name\":\"Vulcan\"}'\n",
       "  id: call_al69pbS0aE2LrEXiYTarQ2ZJ\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[3] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[4] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_al69pbS0aE2LrEXiYTarQ2ZJ\n",
       "\n",
       "VMC</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 199 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_ticker\",\n",
       "      \"description\": \"Get company stock ticker from its name.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"company_name\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"company_name\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_data\",\n",
       "      \"description\": \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"symbol\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"start_date\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"end_date\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"symbol\",\n",
       "          \"start_date\",\n",
       "          \"end_date\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\n",
       "\n",
       "2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\n",
       "\n",
       "3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\n",
       "\n",
       "4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.\n",
       "--\n",
       "assistant: [{'function': {'name': 'get_stock_ticker', 'arguments': '{\"company_name\":\"Vulcan\"}'}, 'id': 'call_al69pbS0aE2LrEXiYTarQ2ZJ', 'type': 'function'}]\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "user: Follow the plan you created to earlier. When you are done, respond to the user.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 0 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=None, role='assistant', tool_calls=[ChatCompletionMessageToolCall(function=Function(arguments='{\"symbol\":\"VMC\",\"start_date\":\"2024-09-01\",\"end_date\":\"2024-09-16\"}', name='get_stock_data'), id='call_s0VrNLHmJksS9Mgvh4tgmXCb', type='function')], function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[5] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_data\n",
       "    arguments: '{\"symbol\":\"VMC\",\"start_date\":\"2024-09-01\",\"end_date\":\"2024-09-16\"}'\n",
       "  id: call_s0VrNLHmJksS9Mgvh4tgmXCb\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[6] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[7] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_s0VrNLHmJksS9Mgvh4tgmXCb\n",
       "\n",
       "[['2024-09-03', 239.02000427246094], ['2024-09-04', 234.8300018310547], ['2024-09-05', 232.16000366210938], ['2024-09-06', 231.97000122070312], ['2024-09-09', 231.8300018310547], ['2024-09-10', 232.9499969482422], ['2024-09-11', 232.9499969482422], ['2024-09-12', 236.27000427246094], ['2024-09-13', 237.47000122070312]]</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 378 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_ticker\",\n",
       "      \"description\": \"Get company stock ticker from its name.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"company_name\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"company_name\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_data\",\n",
       "      \"description\": \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"symbol\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"start_date\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"end_date\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"symbol\",\n",
       "          \"start_date\",\n",
       "          \"end_date\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\n",
       "\n",
       "2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\n",
       "\n",
       "3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\n",
       "\n",
       "4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.\n",
       "--\n",
       "assistant: [{'function': {'name': 'get_stock_ticker', 'arguments': '{\"company_name\":\"Vulcan\"}'}, 'id': 'call_al69pbS0aE2LrEXiYTarQ2ZJ', 'type': 'function'}]\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "assistant: [{'function': {'name': 'get_stock_data', 'arguments': '{\"symbol\":\"VMC\",\"start_date\":\"2024-09-01\",\"end_date\":\"2024-09-16\"}'}, 'id': 'call_s0VrNLHmJksS9Mgvh4tgmXCb', 'type': 'function'}]\n",
       "--\n",
       "tool: [['2024-09-03', 239.02000427246094], ['2024-09-04', 234.8300018310547], ['2024-09-05', 232.16000366210938], ['2024-09-06', 231.97000122070312], ['2024-09-09', 231.8300018310547], ['2024-09-10', 232.9499969482422], ['2024-09-11', 232.9499969482422], ['2024-09-12', 236.27000427246094], ['2024-09-13', 237.47000122070312]]\n",
       "--\n",
       "user: Follow the plan you created to earlier. When you are done, respond to the user.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 87 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=\"Vulcan Materials Company, trading under the stock ticker symbol VMC, is a major producer of construction aggregates and a leading producer of other construction materials. Recently, its stock prices have shown some fluctuations, with prices ranging from approximately $231.83 to $239.02 between September 1 and September 13, 2024. This reflects a stable yet slightly volatile market performance, indicative of the construction industry's dynamics and economic conditions.\", role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[8] Assistant</h4><pre class='step-text'>kind: assistant\n",
       "\n",
       "Vulcan Materials Company, trading under the stock ticker symbol VMC, is a major producer of construction aggregates and a leading producer of other construction materials. Recently, its stock prices have shown some fluctuations, with prices ranging from approximately $231.83 to $239.02 between September 1 and September 13, 2024. This reflects a stable yet slightly volatile market performance, indicative of the construction industry's dynamics and economic conditions.</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[9] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: plan\n",
       "</pre></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from tapeagents.core import SetNextNode\n",
    "from tapeagents.dialog_tape import AssistantThought, ToolCalls\n",
    "from tapeagents.environment import ToolEnvironment\n",
    "from tapeagents.orchestrator import main_loop\n",
    "from tapeagents.tools.stock import get_stock_data, get_stock_ticker\n",
    "\n",
    "system_instruction = f\"\"\"\n",
    "You will help the user to learn about financials of companies.\n",
    "Use as many relevant tools as possible to include more details and facts in your responses.\n",
    "Today is {today}.\n",
    "\"\"\"\n",
    "system_message = {\"role\": \"system\", \"content\": system_instruction}\n",
    "\n",
    "env = ToolEnvironment([get_stock_ticker, get_stock_data])\n",
    "\n",
    "\n",
    "class PlanNode(Node):\n",
    "    name: str = \"plan\"\n",
    "\n",
    "    def make_prompt(self, agent, tape: DialogTape) -> Prompt:\n",
    "        guidance = \"Write a natural language plan on how to use tools help the user. Output a list of numbered items, like 1., 2., 3., etc.\"\n",
    "        guidance_message = {\"role\": \"user\", \"content\": guidance}\n",
    "        return Prompt(\n",
    "            messages=[system_message] + tape_to_messages(tape) + [guidance_message], tools=env.get_tool_schema_dicts()\n",
    "        )\n",
    "\n",
    "    def generate_steps(self, agent, tape, llm_stream: LLMStream):\n",
    "        if content := llm_stream.get_output().content:\n",
    "            yield AssistantThought(content=content)\n",
    "        else:\n",
    "            raise ValueError()\n",
    "\n",
    "\n",
    "class ActNode(Node):\n",
    "    name: str = \"act\"\n",
    "\n",
    "    def make_prompt(self, agent, tape: DialogTape) -> Prompt:\n",
    "        guidance = \"Follow the plan you created to earlier. When you are done, respond to the user.\"\n",
    "        guidance_message = {\"role\": \"user\", \"content\": guidance}\n",
    "        return Prompt(\n",
    "            messages=[system_message] + tape_to_messages(tape) + [guidance_message], tools=env.get_tool_schema_dicts()\n",
    "        )\n",
    "\n",
    "    def generate_steps(self, agent, tape, llm_stream: LLMStream):\n",
    "        o = llm_stream.get_output()\n",
    "        if o.content:\n",
    "            yield AssistantStep(content=o.content)\n",
    "            yield SetNextNode(next_node=\"plan\")\n",
    "        elif o.tool_calls:\n",
    "            yield ToolCalls.from_llm_output(o)\n",
    "            yield SetNextNode(next_node=\"act\")\n",
    "        else:\n",
    "            raise ValueError()\n",
    "\n",
    "\n",
    "agent1 = Agent.create(LiteLLM(model_name=\"gpt-4o\", parameters={\"temperature\": 0.1}), nodes=[PlanNode(), ActNode()])\n",
    "\n",
    "print(\"Run the agent!\")\n",
    "final_tape1 = None\n",
    "for event in main_loop(agent1, DialogTape() + [UserStep(content=\"Tell me about Vulcan in 3 sentences\")], env):\n",
    "    if ae := event.agent_event:\n",
    "        if ae.step:\n",
    "            print(ae.step.model_dump_json(indent=2))\n",
    "        if ae.final_tape:\n",
    "            final_tape1 = ae.final_tape\n",
    "    if event.observation:\n",
    "        print(event.observation.model_dump_json(indent=2))\n",
    "assert final_tape1\n",
    "print(\"Final tape:\")\n",
    "HTML(render_tape_with_prompts(final_tape1, PrettyRenderer()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The main new thing in this example is the environment. In TapeAgents framework the environment responds to the agent `Action` steps with `Observation` steps. We expect you to use the environment to encapsulate tool use, retrieval, code execution: everything that is non-deterministic, non-stationary, or computationally heavy. On the contrary, we encourage you to implements the agent's deterministic decision-making in `make_prompt` and `generate_steps` methods.\n",
    "\n",
    "Here we use a pre-defined `main_loop` orchestrator to run the agent and the environment. `main_loop` is a generator of events that you can use as you wish. You are free to implement your own orchestration paradigm with a fine-grained control over what actions get to be executed."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. Agent configuration, resumption"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's try building a similar agent with an open-weights LLAMA3 70B models. Conveniently, [Together AI](together.ai) offers API endpoints. You can create an account and get API key with some free quota."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Access to Hugging Face Gated Models\n",
    "Make sure you have access to each model we are going to use (read more [here](https://huggingface.co/docs/hub/en/models-gated#access-gated-models-as-a-user)):\n",
    "- https://huggingface.co/meta-llama/Meta-Llama-3-70B-Instruct\n",
    "\n",
    "##### Troubleshoot\n",
    "- If you receive a 401 error, it means the authentication failed.\n",
    "- If you receive a 403 error, it means you are authenticated but not authorized to access the specific model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:47.736220Z",
     "iopub.status.busy": "2025-02-24T13:10:47.735863Z",
     "iopub.status.idle": "2025-02-24T13:10:47.740633Z",
     "shell.execute_reply": "2025-02-24T13:10:47.740020Z"
    }
   },
   "outputs": [],
   "source": [
    "if \"TAPEAGENTS_LLM_TOKEN\" not in os.environ:\n",
    "    os.environ[\"TAPEAGENTS_LLM_TOKEN\"] = \"<your-together-ai-api-key>\"\n",
    "\n",
    "if \"HF_TOKEN\" not in os.environ:\n",
    "    # We need this to acces the model's tokenizer\n",
    "    os.environ[\"HF_TOKEN\"] = \"<your-hugging-face-api-key>\"\n",
    "\n",
    "if \"SERPER_API_KEY\" not in os.environ:  # web search, api key for https://serper.dev\n",
    "    os.environ[\"SERPER_API_KEY\"] = \"<your-serper-api-key>\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.1. Setup your Agents\n",
    "\n",
    "We've found that LLAMA3 function-calling is not yet battle-ready. We will use the structured output approach to make it call tools instead. We are also making this agent trainable by adding `make_llm_output` methods to each node. `Node.make_llm_output` defines how a node can reconstruct the LLM completion message that would be required to make the steps from the given tape at the given index. You can think of `Node.make_llm_output` as the inverse of `Node.generate_steps`.\n",
    "\n",
    "When you run the code below, you might see a different behavior every time. Often the LLAMA-based agent gets stuck in a loop. We will look into how TapeAgents supports you in addressing this issue by\n",
    "- tuning the prompt and resuming the agent exactly where it got stuck \n",
    "- producing training text from a different agent's tape "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:47.743495Z",
     "iopub.status.busy": "2025-02-24T13:10:47.743214Z",
     "iopub.status.idle": "2025-02-24T13:10:47.766444Z",
     "shell.execute_reply": "2025-02-24T13:10:47.766099Z"
    }
   },
   "outputs": [],
   "source": [
    "import json\n",
    "\n",
    "from tapeagents.llms import TrainableLLM\n",
    "from tapeagents.prompting import step_to_message\n",
    "from tapeagents.tool_calling import FunctionCall, ToolCall\n",
    "\n",
    "env = ToolEnvironment([get_stock_ticker, get_stock_data])\n",
    "\n",
    "system_instruction = f\"\"\"\n",
    "You will help the user to learn about financials of companies.\n",
    "Use as many relevant tools as possible to include more details and facts in your responses.\n",
    "Today is {today}.\n",
    "\n",
    "You have access to the following tools: {env.get_tool_schema_dicts()}\"\"\"\n",
    "\n",
    "planning_guidance = \"Write a natural language plan on how to use tools help the user. Output a list of numbered items, like 1., 2., 3., etc.\"\n",
    "\n",
    "call_or_respond_guidance = \"\"\"\n",
    "Follow the plan you created earlier. When you are done, respond to the user.\n",
    "If you want to call a or several tools, output JSON like this\n",
    "{\"kind\": \"tool_call\", \"tool_name\": \"...\", \"parameters\": \"... unquoted parameters json ...\"}\n",
    "If you have called all the tools in the plan, respond to the user with the JSON of the form\n",
    "{\"kind\": \"response\", \"content\": \"... you response ... \"}.\n",
    "Output ONE JSON OBJECT ONLY PER LINE ONLY AND NOTHING ELSE.\n",
    "\"\"\"\n",
    "\n",
    "\n",
    "class PlanNode(Node):\n",
    "    name: str = \"plan\"\n",
    "\n",
    "    def make_prompt(self, agent, tape: DialogTape) -> Prompt:\n",
    "        system_message = {\"role\": \"system\", \"content\": system_instruction}\n",
    "        guidance_message = {\"role\": \"user\", \"content\": agent.templates[\"planning\"]}\n",
    "        return Prompt(messages=[system_message] + tape_to_messages(tape) + [guidance_message])\n",
    "\n",
    "    def generate_steps(self, agent, tape, llm_stream: LLMStream):\n",
    "        if content := getattr(llm_stream.get_output(), \"content\", None):\n",
    "            yield AssistantThought(content=content)\n",
    "        else:\n",
    "            raise ValueError()\n",
    "\n",
    "    def make_llm_output(self, agent, tape: DialogTape, index: int) -> LLMOutput:\n",
    "        if not isinstance(current := tape[index], AssistantThought):\n",
    "            raise ValueError()\n",
    "        return LLMOutput(role=\"assistant\", content=current.content)\n",
    "\n",
    "\n",
    "class ActNode(Node):\n",
    "    name: str = \"act\"\n",
    "\n",
    "    def make_prompt(self, agent, tape: DialogTape) -> Prompt:\n",
    "        system_message = {\"role\": \"system\", \"content\": system_instruction}\n",
    "        guidance_message = {\"role\": \"user\", \"content\": agent.templates[\"call_or_respond\"]}\n",
    "        messages = [system_message]\n",
    "        for step in tape:\n",
    "            if isinstance(step, (ToolCalls)):\n",
    "                messages.append({\"role\": \"assistant\", \"content\": _llm_message_content(step)})\n",
    "            elif not isinstance(step, SetNextNode):\n",
    "                messages.append(step_to_message(step))\n",
    "        messages += [guidance_message]\n",
    "        return Prompt(messages=messages)\n",
    "\n",
    "    def generate_steps(self, agent, tape, llm_stream: LLMStream):\n",
    "        o = llm_stream.get_output()\n",
    "        try:\n",
    "            assert o.content\n",
    "            tool_calls = []\n",
    "            response = None\n",
    "            for line in o.content.split(\"\\n\"):\n",
    "                data = json.loads(line)\n",
    "                if data.get(\"kind\") == \"response\":\n",
    "                    response = data[\"content\"]\n",
    "                elif data.get(\"kind\") == \"tool_call\":\n",
    "                    tool_call = ToolCall(\n",
    "                        function=FunctionCall(name=data[\"tool_name\"], arguments=json.dumps(data[\"parameters\"])),\n",
    "                        # tool call must be a unique string, it helps to make it something deterministic\n",
    "                        id=f\"tool_call_{len(tool_calls)}_node_starts_at_{len(tape)}\",\n",
    "                    )\n",
    "                    tool_calls.append(tool_call)\n",
    "                else:\n",
    "                    yield AssistantStep(content=\"Invalid LLM output: kind field must be 'response' or 'tool_call'\")\n",
    "            if response and tool_calls:\n",
    "                yield AssistantStep(content=\"Invalid LLM output: response and tool_call cannot be in the same message\")\n",
    "            if response:\n",
    "                yield AssistantStep(content=response)\n",
    "                yield SetNextNode(next_node=\"plan\")\n",
    "            if tool_calls:\n",
    "                yield ToolCalls(tool_calls=tool_calls)\n",
    "                yield SetNextNode(next_node=\"act\")\n",
    "        except Exception as e:\n",
    "            yield AssistantStep(content=\"Invalid JSON object: \" + str(e))\n",
    "\n",
    "    def make_llm_output(self, agent, tape: DialogTape, index: int) -> LLMOutput:\n",
    "        if not isinstance(step := tape[index], AssistantStep | ToolCalls):\n",
    "            raise ValueError()\n",
    "        content = _llm_message_content(step)\n",
    "        return LLMOutput(role=\"assistant\", content=content)\n",
    "\n",
    "\n",
    "def _llm_message_content(step: AssistantStep | ToolCalls):\n",
    "    \"\"\"Helper function to make both the prompt and the target completion\"\"\"\n",
    "    match step:\n",
    "        case AssistantStep():\n",
    "            return json.dumps({\"kind\": \"response\", \"content\": step.content})\n",
    "        case ToolCalls():\n",
    "            content = \"\"\n",
    "            for tool_call in step.tool_calls:\n",
    "                if content:\n",
    "                    content += \"\\n\"\n",
    "                content += json.dumps(\n",
    "                    {\n",
    "                        \"kind\": \"tool_call\",\n",
    "                        \"tool_name\": tool_call.function.name,\n",
    "                        \"parameters\": json.loads(tool_call.function.arguments),\n",
    "                    }\n",
    "                )\n",
    "            return content\n",
    "        case _:\n",
    "            raise ValueError()\n",
    "\n",
    "\n",
    "agent2 = Agent.create(\n",
    "    TrainableLLM(\n",
    "        base_url=\"https://api.together.xyz\",\n",
    "        model_name=\"meta-llama/Meta-Llama-3-70B-Instruct-Turbo\",\n",
    "        tokenizer_name=\"meta-llama/Meta-Llama-3-70B-Instruct\",\n",
    "        parameters=dict(temperature=0.01),\n",
    "    ),\n",
    "    templates={\n",
    "        \"system\": system_instruction,\n",
    "        \"planning\": planning_guidance,\n",
    "        \"call_or_respond\": call_or_respond_guidance,\n",
    "    },\n",
    "    nodes=[PlanNode(), ActNode()],\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.2. Run your Agent"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:10:47.767951Z",
     "iopub.status.busy": "2025-02-24T13:10:47.767836Z",
     "iopub.status.idle": "2025-02-24T13:11:12.985986Z",
     "shell.execute_reply": "2025-02-24T13:11:12.985258Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Run LLAMA agent!\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"e6d828d8-c4e9-4ea5-bb68-e21753fde2bc\",\n",
      "    \"prompt_id\": \"76e354fb-1ca2-4dde-948e-a57de9d994cc\",\n",
      "    \"node\": \"plan\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"assistant_thought\",\n",
      "  \"content\": \"Here's a plan on how to use the provided tools to help the user learn about Vulcan Materials' financials:\\n\\n1. First, use the `get_stock_ticker` function to retrieve the stock ticker symbol for Vulcan Materials by passing \\\"Vulcan Materials\\\" as the `company_name` parameter.\\n\\n2. Once the stock ticker symbol is obtained, use the `get_stock_data` function to retrieve the historical stock prices for Vulcan Materials over a specified date range, such as the past year or past quarter.\\n\\n3. Analyze the retrieved stock price data to identify trends, patterns, and key events that may have impacted the company's financial performance.\\n\\n4. Provide the user with a summary of the company's financial performance, including its current stock price, historical price trends, and any notable events or announcements that may have affected its stock price.\\n\\n5. Offer additional insights or recommendations based on the analyzed data, such as whether the company's stock appears to be undervalued or overvalued, or whether there are any potential risks or opportunities that investors should be aware of.\\n\\n6. If the user requests more information, use the `get_stock_data` function to retrieve additional data, such as the company's dividend history, earnings per share, or other relevant financial metrics.\\n\\n7. Provide the user with visualizations or charts to help illustrate the company's financial performance and trends, making it easier to understand and interpret the data.\\n\\nBy following this plan, we can leverage the provided tools to provide the user with a comprehensive overview of Vulcan Materials' financials and help them make more informed investment decisions.\"\n",
      "}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"f8abee98-b784-4879-b414-58865305650c\",\n",
      "    \"prompt_id\": \"356dde09-0f44-442f-86b5-c4de8c7fe1a3\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"assistant\",\n",
      "  \"tool_calls\": [\n",
      "    {\n",
      "      \"function\": {\n",
      "        \"name\": \"get_stock_ticker\",\n",
      "        \"arguments\": \"{\\\"company_name\\\": \\\"Vulcan Materials\\\"}\"\n",
      "      },\n",
      "      \"id\": \"tool_call_0_node_starts_at_2\",\n",
      "      \"type\": \"function\"\n",
      "    }\n",
      "  ]\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"9e7ada9c-c25e-49d2-9c08-170df0a36fe4\",\n",
      "    \"prompt_id\": \"356dde09-0f44-442f-86b5-c4de8c7fe1a3\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"set_next_node\",\n",
      "  \"next_node\": \"act\"\n",
      "}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"0da14f6c-bccf-4845-9894-60fe16f70928\",\n",
      "    \"prompt_id\": \"\",\n",
      "    \"node\": \"\",\n",
      "    \"agent\": \"\",\n",
      "    \"llm\": \"\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"tool\",\n",
      "  \"content\": \"VMC\",\n",
      "  \"tool_call_id\": \"tool_call_0_node_starts_at_2\"\n",
      "}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"b5cc755c-d09b-4fcc-89f4-4a6772571979\",\n",
      "    \"prompt_id\": \"132a603e-a882-44de-a241-91b7969ba824\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"assistant\",\n",
      "  \"tool_calls\": [\n",
      "    {\n",
      "      \"function\": {\n",
      "        \"name\": \"get_stock_data\",\n",
      "        \"arguments\": \"{\\\"symbol\\\": \\\"VMC\\\", \\\"start_date\\\": \\\"2023-09-17\\\", \\\"end_date\\\": \\\"2024-09-17\\\"}\"\n",
      "      },\n",
      "      \"id\": \"tool_call_0_node_starts_at_5\",\n",
      "      \"type\": \"function\"\n",
      "    }\n",
      "  ]\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"996bebbc-6ef2-4fee-ae23-b0e49a8fc226\",\n",
      "    \"prompt_id\": \"132a603e-a882-44de-a241-91b7969ba824\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"set_next_node\",\n",
      "  \"next_node\": \"act\"\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"ff56fd07-6fb1-45e5-a4b3-19887024b60f\",\n",
      "    \"prompt_id\": \"\",\n",
      "    \"node\": \"\",\n",
      "    \"agent\": \"\",\n",
      "    \"llm\": \"\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"tool\",\n",
      "  \"content\": [\n",
      "    [\n",
      "      \"2023-09-18\",\n",
      "      211.69000244140625\n",
      "    ],\n",
      "    [\n",
      "      \"2023-09-22\",\n",
      "      200.6199951171875\n",
      "    ],\n",
      "    [\n",
      "      \"2023-09-28\",\n",
      "      205.02999877929688\n",
      "    ],\n",
      "    [\n",
      "      \"2023-10-04\",\n",
      "      205.02999877929688\n",
      "    ],\n",
      "    [\n",
      "      \"2023-10-10\",\n",
      "      211.0\n",
      "    ],\n",
      "    [\n",
      "      \"2023-10-16\",\n",
      "      213.0399932861328\n",
      "    ],\n",
      "    [\n",
      "      \"2023-10-20\",\n",
      "      201.5500030517578\n",
      "    ],\n",
      "    [\n",
      "      \"2023-10-26\",\n",
      "      193.97000122070312\n",
      "    ],\n",
      "    [\n",
      "      \"2023-11-01\",\n",
      "      203.8000030517578\n",
      "    ],\n",
      "    [\n",
      "      \"2023-11-07\",\n",
      "      207.2100067138672\n",
      "    ],\n",
      "    [\n",
      "      \"2023-11-13\",\n",
      "      210.49000549316406\n",
      "    ],\n",
      "    [\n",
      "      \"2023-11-17\",\n",
      "      212.35000610351562\n",
      "    ],\n",
      "    [\n",
      "      \"2023-11-24\",\n",
      "      211.67999267578125\n",
      "    ],\n",
      "    [\n",
      "      \"2023-11-30\",\n",
      "      213.55999755859375\n",
      "    ],\n",
      "    [\n",
      "      \"2023-12-06\",\n",
      "      211.92999267578125\n",
      "    ],\n",
      "    [\n",
      "      \"2023-12-12\",\n",
      "      220.89999389648438\n",
      "    ],\n",
      "    [\n",
      "      \"2023-12-18\",\n",
      "      222.6999969482422\n",
      "    ],\n",
      "    [\n",
      "      \"2023-12-22\",\n",
      "      224.92999267578125\n",
      "    ],\n",
      "    [\n",
      "      \"2023-12-29\",\n",
      "      227.00999450683594\n",
      "    ],\n",
      "    [\n",
      "      \"2024-01-05\",\n",
      "      221.6199951171875\n",
      "    ],\n",
      "    [\n",
      "      \"2024-01-11\",\n",
      "      224.36000061035156\n",
      "    ],\n",
      "    [\n",
      "      \"2024-01-18\",\n",
      "      225.1199951171875\n",
      "    ],\n",
      "    [\n",
      "      \"2024-01-24\",\n",
      "      219.8000030517578\n",
      "    ],\n",
      "    [\n",
      "      \"2024-01-30\",\n",
      "      231.0500030517578\n",
      "    ],\n",
      "    [\n",
      "      \"2024-02-05\",\n",
      "      229.64999389648438\n",
      "    ],\n",
      "    [\n",
      "      \"2024-02-09\",\n",
      "      240.0\n",
      "    ],\n",
      "    [\n",
      "      \"2024-02-15\",\n",
      "      242.4600067138672\n",
      "    ],\n",
      "    [\n",
      "      \"2024-02-22\",\n",
      "      256.94000244140625\n",
      "    ],\n",
      "    [\n",
      "      \"2024-02-28\",\n",
      "      262.29998779296875\n",
      "    ],\n",
      "    [\n",
      "      \"2024-03-05\",\n",
      "      264.9800109863281\n",
      "    ],\n",
      "    [\n",
      "      \"2024-03-11\",\n",
      "      264.95001220703125\n",
      "    ],\n",
      "    [\n",
      "      \"2024-03-15\",\n",
      "      266.8599853515625\n",
      "    ],\n",
      "    [\n",
      "      \"2024-03-21\",\n",
      "      275.5899963378906\n",
      "    ],\n",
      "    [\n",
      "      \"2024-03-27\",\n",
      "      272.7900085449219\n",
      "    ],\n",
      "    [\n",
      "      \"2024-04-03\",\n",
      "      268.7699890136719\n",
      "    ],\n",
      "    [\n",
      "      \"2024-04-09\",\n",
      "      265.6199951171875\n",
      "    ],\n",
      "    [\n",
      "      \"2024-04-15\",\n",
      "      260.4700012207031\n",
      "    ],\n",
      "    [\n",
      "      \"2024-04-19\",\n",
      "      252.05999755859375\n",
      "    ],\n",
      "    [\n",
      "      \"2024-04-25\",\n",
      "      258.5400085449219\n",
      "    ],\n",
      "    [\n",
      "      \"2024-05-01\",\n",
      "      259.7300109863281\n",
      "    ],\n",
      "    [\n",
      "      \"2024-05-07\",\n",
      "      268.3500061035156\n",
      "    ],\n",
      "    [\n",
      "      \"2024-05-13\",\n",
      "      270.0899963378906\n",
      "    ],\n",
      "    [\n",
      "      \"2024-05-17\",\n",
      "      259.1000061035156\n",
      "    ],\n",
      "    [\n",
      "      \"2024-05-23\",\n",
      "      257.2099914550781\n",
      "    ],\n",
      "    [\n",
      "      \"2024-05-30\",\n",
      "      254.0800018310547\n",
      "    ],\n",
      "    [\n",
      "      \"2024-06-05\",\n",
      "      251.17999267578125\n",
      "    ],\n",
      "    [\n",
      "      \"2024-06-11\",\n",
      "      244.63999938964844\n",
      "    ],\n",
      "    [\n",
      "      \"2024-06-17\",\n",
      "      256.3500061035156\n",
      "    ],\n",
      "    [\n",
      "      \"2024-06-24\",\n",
      "      247.5800018310547\n",
      "    ],\n",
      "    [\n",
      "      \"2024-06-28\",\n",
      "      248.67999267578125\n",
      "    ],\n",
      "    [\n",
      "      \"2024-07-05\",\n",
      "      244.2100067138672\n",
      "    ],\n",
      "    [\n",
      "      \"2024-07-11\",\n",
      "      254.69000244140625\n",
      "    ],\n",
      "    [\n",
      "      \"2024-07-17\",\n",
      "      255.8000030517578\n",
      "    ],\n",
      "    [\n",
      "      \"2024-07-23\",\n",
      "      260.1499938964844\n",
      "    ],\n",
      "    [\n",
      "      \"2024-07-29\",\n",
      "      264.5299987792969\n",
      "    ],\n",
      "    [\n",
      "      \"2024-08-02\",\n",
      "      266.3399963378906\n",
      "    ],\n",
      "    [\n",
      "      \"2024-08-08\",\n",
      "      246.3300018310547\n",
      "    ],\n",
      "    [\n",
      "      \"2024-08-14\",\n",
      "      242.5\n",
      "    ],\n",
      "    [\n",
      "      \"2024-08-20\",\n",
      "      245.60000610351562\n",
      "    ],\n",
      "    [\n",
      "      \"2024-08-26\",\n",
      "      250.67999267578125\n",
      "    ],\n",
      "    [\n",
      "      \"2024-08-30\",\n",
      "      245.2100067138672\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-06\",\n",
      "      231.97000122070312\n",
      "    ],\n",
      "    [\n",
      "      \"2024-09-12\",\n",
      "      236.27000427246094\n",
      "    ]\n",
      "  ],\n",
      "  \"tool_call_id\": \"tool_call_0_node_starts_at_5\"\n",
      "}\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"82cbb965-9285-4f59-bb3c-76afdb01d449\",\n",
      "    \"prompt_id\": \"b24cc994-2020-48a4-a2f6-d3f0eb50d4ff\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"assistant\",\n",
      "  \"content\": \"Vulcan Materials (VMC) is a leading producer of construction materials, including aggregates, asphalt, and ready-mixed concrete. As of September 17, 2024, the company's stock is trading at around 236.27, which is a slight decrease from its recent high. Over the past year, VMC's stock has experienced some volatility, with a high of 275.59 and a low of 193.97. Despite this, the company's financial performance has remained strong, with revenue growth and stable profit margins. Overall, Vulcan Materials appears to be a solid investment opportunity for those looking to invest in the construction materials industry.\"\n",
      "}\n",
      "{\n",
      "  \"metadata\": {\n",
      "    \"id\": \"d00fe01f-ef6a-46a7-95fb-97365b8f9314\",\n",
      "    \"prompt_id\": \"b24cc994-2020-48a4-a2f6-d3f0eb50d4ff\",\n",
      "    \"node\": \"act\",\n",
      "    \"agent\": \"Agent\",\n",
      "    \"llm\": \"default\",\n",
      "    \"other\": {}\n",
      "  },\n",
      "  \"kind\": \"set_next_node\",\n",
      "  \"next_node\": \"plan\"\n",
      "}\n",
      "Final tape:\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<style>.basic-renderer-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; white-space: pre-wrap; color: black;}.basic-prompt-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; color: black;}.episode-row { display: flex; align-items: end; }.agent-column { width: 50%; }.user-column { width: 25%; }.annotator-column { width: 25%; }.inner-tape-container { display: flex }.inner-tape-indent { width: 10%; }.inner-tape { width: 90%; }.agent_node { text-align: center; position: relative; margin: 12px 0; }.agent_node hr { border: 1px solid #e1e1e1; margin: 0 !important;}.agent_node span { position: absolute; right: 0; top: -0.6em; background: white; color: grey !important; padding: 0 10px; }</style><style>.observation { background-color: #baffc9;; }.error_observation { background-color: #dd0000; }.action { background-color: #cccccc; }.thought { background-color: #ffffdb; }.call { background-color: #ffffff; }.respond { background-color: #ffffff; }.step-header { margin: 2pt 2pt 2pt 0 !important; }.step-text { font-size: 12px; white-space: pre-wrap; word-wrap: break-word;}</style><h3 style='margin: 2px'>Metadata</h3><details> <summary>id: ca55fcad-4081-43ae-8ec2-dc122020062f</summary> <div class='basic-renderer-box'>author: Agent\n",
       "author_tape_id: null\n",
       "error: null\n",
       "id: ca55fcad-4081-43ae-8ec2-dc122020062f\n",
       "n_added_steps: 2\n",
       "parent_id: e2edc5a5-c87d-48c6-bfc4-abba796dcabe\n",
       "result: {}\n",
       "</div> </details><h3 style='margin: 2px'>Steps</h3><div class='basic-renderer-box observation'><h4 class='step-header'>[0] User</h4><pre class='step-text'>kind: user\n",
       "\n",
       "Tell me about Vulcan Materials in 3 sentences</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 341 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "You have access to the following tools: [{'type': 'function', 'function': {'name': 'get_stock_ticker', 'description': 'Get company stock ticker from its name.', 'parameters': {'properties': {'company_name': {'type': 'string'}}, 'required': ['company_name'], 'type': 'object'}}}, {'type': 'function', 'function': {'name': 'get_stock_data', 'description': \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\", 'parameters': {'properties': {'symbol': {'type': 'string'}, 'start_date': {'type': 'string'}, 'end_date': {'type': 'string'}}, 'required': ['symbol', 'start_date', 'end_date'], 'type': 'object'}}}]\n",
       "--\n",
       "user: Tell me about Vulcan Materials in 3 sentences\n",
       "--\n",
       "user: Write a natural language plan on how to use tools help the user. Output a list of numbered items, like 1., 2., 3., etc.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 323 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='Here\\'s a plan on how to use the provided tools to help the user learn about Vulcan Materials\\' financials:\\n\\n1. First, use the `get_stock_ticker` function to retrieve the stock ticker symbol for Vulcan Materials by passing \"Vulcan Materials\" as the `company_name` parameter.\\n\\n2. Once the stock ticker symbol is obtained, use the `get_stock_data` function to retrieve the historical stock prices for Vulcan Materials over a specified date range, such as the past year or past quarter.\\n\\n3. Analyze the retrieved stock price data to identify trends, patterns, and key events that may have impacted the company\\'s financial performance.\\n\\n4. Provide the user with a summary of the company\\'s financial performance, including its current stock price, historical price trends, and any notable events or announcements that may have affected its stock price.\\n\\n5. Offer additional insights or recommendations based on the analyzed data, such as whether the company\\'s stock appears to be undervalued or overvalued, or whether there are any potential risks or opportunities that investors should be aware of.\\n\\n6. If the user requests more information, use the `get_stock_data` function to retrieve additional data, such as the company\\'s dividend history, earnings per share, or other relevant financial metrics.\\n\\n7. Provide the user with visualizations or charts to help illustrate the company\\'s financial performance and trends, making it easier to understand and interpret the data.\\n\\nBy following this plan, we can leverage the provided tools to provide the user with a comprehensive overview of Vulcan Materials\\' financials and help them make more informed investment decisions.', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box thought'><h4 class='step-header'>[1] Thought: AssistantThought</h4><pre class='step-text'>kind: assistant_thought\n",
       "\n",
       "<details><summary>1665 characters ...</summary>Here's a plan on how to use the provided tools to help the user learn about Vulcan Materials' financials:\n",
       "\n",
       "1. First, use the `get_stock_ticker` function to retrieve the stock ticker symbol for Vulcan Materials by passing \"Vulcan Materials\" as the `company_name` parameter.\n",
       "\n",
       "2. Once the stock ticker symbol is obtained, use the `get_stock_data` function to retrieve the historical stock prices for Vulcan Materials over a specified date range, such as the past year or past quarter.\n",
       "\n",
       "3. Analyze the retrieved stock price data to identify trends, patterns, and key events that may have impacted the company's financial performance.\n",
       "\n",
       "4. Provide the user with a summary of the company's financial performance, including its current stock price, historical price trends, and any notable events or announcements that may have affected its stock price.\n",
       "\n",
       "5. Offer additional insights or recommendations based on the analyzed data, such as whether the company's stock appears to be undervalued or overvalued, or whether there are any potential risks or opportunities that investors should be aware of.\n",
       "\n",
       "6. If the user requests more information, use the `get_stock_data` function to retrieve additional data, such as the company's dividend history, earnings per share, or other relevant financial metrics.\n",
       "\n",
       "7. Provide the user with visualizations or charts to help illustrate the company's financial performance and trends, making it easier to understand and interpret the data.\n",
       "\n",
       "By following this plan, we can leverage the provided tools to provide the user with a comprehensive overview of Vulcan Materials' financials and help them make more informed investment decisions.</details></pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 739 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "You have access to the following tools: [{'type': 'function', 'function': {'name': 'get_stock_ticker', 'description': 'Get company stock ticker from its name.', 'parameters': {'properties': {'company_name': {'type': 'string'}}, 'required': ['company_name'], 'type': 'object'}}}, {'type': 'function', 'function': {'name': 'get_stock_data', 'description': \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\", 'parameters': {'properties': {'symbol': {'type': 'string'}, 'start_date': {'type': 'string'}, 'end_date': {'type': 'string'}}, 'required': ['symbol', 'start_date', 'end_date'], 'type': 'object'}}}]\n",
       "--\n",
       "user: Tell me about Vulcan Materials in 3 sentences\n",
       "--\n",
       "assistant: Here's a plan on how to use the provided tools to help the user learn about Vulcan Materials' financials:\n",
       "\n",
       "1. First, use the `get_stock_ticker` function to retrieve the stock ticker symbol for Vulcan Materials by passing \"Vulcan Materials\" as the `company_name` parameter.\n",
       "\n",
       "2. Once the stock ticker symbol is obtained, use the `get_stock_data` function to retrieve the historical stock prices for Vulcan Materials over a specified date range, such as the past year or past quarter.\n",
       "\n",
       "3. Analyze the retrieved stock price data to identify trends, patterns, and key events that may have impacted the company's financial performance.\n",
       "\n",
       "4. Provide the user with a summary of the company's financial performance, including its current stock price, historical price trends, and any notable events or announcements that may have affected its stock price.\n",
       "\n",
       "5. Offer additional insights or recommendations based on the analyzed data, such as whether the company's stock appears to be undervalued or overvalued, or whether there are any potential risks or opportunities that investors should be aware of.\n",
       "\n",
       "6. If the user requests more information, use the `get_stock_data` function to retrieve additional data, such as the company's dividend history, earnings per share, or other relevant financial metrics.\n",
       "\n",
       "7. Provide the user with visualizations or charts to help illustrate the company's financial performance and trends, making it easier to understand and interpret the data.\n",
       "\n",
       "By following this plan, we can leverage the provided tools to provide the user with a comprehensive overview of Vulcan Materials' financials and help them make more informed investment decisions.\n",
       "--\n",
       "user: \n",
       "Follow the plan you created earlier. When you are done, respond to the user.\n",
       "If you want to call a or several tools, output JSON like this\n",
       "{\"kind\": \"tool_call\", \"tool_name\": \"...\", \"parameters\": \"... unquoted parameters json ...\"}\n",
       "If you have called all the tools in the plan, respond to the user with the JSON of the form\n",
       "{\"kind\": \"response\", \"content\": \"... you response ... \"}.\n",
       "Output ONE JSON OBJECT ONLY PER LINE ONLY AND NOTHING ELSE.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 31 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='{\"kind\": \"tool_call\", \"tool_name\": \"get_stock_ticker\", \"parameters\": {\"company_name\": \"Vulcan Materials\"}}', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[2] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_ticker\n",
       "    arguments: '{\"company_name\": \"Vulcan Materials\"}'\n",
       "  id: tool_call_0_node_starts_at_2\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[3] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[4] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: tool_call_0_node_starts_at_2\n",
       "\n",
       "VMC</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 781 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "You have access to the following tools: [{'type': 'function', 'function': {'name': 'get_stock_ticker', 'description': 'Get company stock ticker from its name.', 'parameters': {'properties': {'company_name': {'type': 'string'}}, 'required': ['company_name'], 'type': 'object'}}}, {'type': 'function', 'function': {'name': 'get_stock_data', 'description': \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\", 'parameters': {'properties': {'symbol': {'type': 'string'}, 'start_date': {'type': 'string'}, 'end_date': {'type': 'string'}}, 'required': ['symbol', 'start_date', 'end_date'], 'type': 'object'}}}]\n",
       "--\n",
       "user: Tell me about Vulcan Materials in 3 sentences\n",
       "--\n",
       "assistant: Here's a plan on how to use the provided tools to help the user learn about Vulcan Materials' financials:\n",
       "\n",
       "1. First, use the `get_stock_ticker` function to retrieve the stock ticker symbol for Vulcan Materials by passing \"Vulcan Materials\" as the `company_name` parameter.\n",
       "\n",
       "2. Once the stock ticker symbol is obtained, use the `get_stock_data` function to retrieve the historical stock prices for Vulcan Materials over a specified date range, such as the past year or past quarter.\n",
       "\n",
       "3. Analyze the retrieved stock price data to identify trends, patterns, and key events that may have impacted the company's financial performance.\n",
       "\n",
       "4. Provide the user with a summary of the company's financial performance, including its current stock price, historical price trends, and any notable events or announcements that may have affected its stock price.\n",
       "\n",
       "5. Offer additional insights or recommendations based on the analyzed data, such as whether the company's stock appears to be undervalued or overvalued, or whether there are any potential risks or opportunities that investors should be aware of.\n",
       "\n",
       "6. If the user requests more information, use the `get_stock_data` function to retrieve additional data, such as the company's dividend history, earnings per share, or other relevant financial metrics.\n",
       "\n",
       "7. Provide the user with visualizations or charts to help illustrate the company's financial performance and trends, making it easier to understand and interpret the data.\n",
       "\n",
       "By following this plan, we can leverage the provided tools to provide the user with a comprehensive overview of Vulcan Materials' financials and help them make more informed investment decisions.\n",
       "--\n",
       "assistant: {\"kind\": \"tool_call\", \"tool_name\": \"get_stock_ticker\", \"parameters\": {\"company_name\": \"Vulcan Materials\"}}\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "user: \n",
       "Follow the plan you created earlier. When you are done, respond to the user.\n",
       "If you want to call a or several tools, output JSON like this\n",
       "{\"kind\": \"tool_call\", \"tool_name\": \"...\", \"parameters\": \"... unquoted parameters json ...\"}\n",
       "If you have called all the tools in the plan, respond to the user with the JSON of the form\n",
       "{\"kind\": \"response\", \"content\": \"... you response ... \"}.\n",
       "Output ONE JSON OBJECT ONLY PER LINE ONLY AND NOTHING ELSE.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 51 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='{\"kind\": \"tool_call\", \"tool_name\": \"get_stock_data\", \"parameters\": {\"symbol\": \"VMC\", \"start_date\": \"2023-09-17\", \"end_date\": \"2024-09-17\"}}', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[5] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_data\n",
       "    arguments: '{\"symbol\": \"VMC\", \"start_date\": \"2023-09-17\", \"end_date\": \"2024-09-17\"}'\n",
       "  id: tool_call_0_node_starts_at_5\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[6] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[7] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: tool_call_0_node_starts_at_5\n",
       "\n",
       "<details><summary>2195 characters ...</summary>[['2023-09-18', 211.69000244140625], ['2023-09-22', 200.6199951171875], ['2023-09-28', 205.02999877929688], ['2023-10-04', 205.02999877929688], ['2023-10-10', 211.0], ['2023-10-16', 213.0399932861328], ['2023-10-20', 201.5500030517578], ['2023-10-26', 193.97000122070312], ['2023-11-01', 203.8000030517578], ['2023-11-07', 207.2100067138672], ['2023-11-13', 210.49000549316406], ['2023-11-17', 212.35000610351562], ['2023-11-24', 211.67999267578125], ['2023-11-30', 213.55999755859375], ['2023-12-06', 211.92999267578125], ['2023-12-12', 220.89999389648438], ['2023-12-18', 222.6999969482422], ['2023-12-22', 224.92999267578125], ['2023-12-29', 227.00999450683594], ['2024-01-05', 221.6199951171875], ['2024-01-11', 224.36000061035156], ['2024-01-18', 225.1199951171875], ['2024-01-24', 219.8000030517578], ['2024-01-30', 231.0500030517578], ['2024-02-05', 229.64999389648438], ['2024-02-09', 240.0], ['2024-02-15', 242.4600067138672], ['2024-02-22', 256.94000244140625], ['2024-02-28', 262.29998779296875], ['2024-03-05', 264.9800109863281], ['2024-03-11', 264.95001220703125], ['2024-03-15', 266.8599853515625], ['2024-03-21', 275.5899963378906], ['2024-03-27', 272.7900085449219], ['2024-04-03', 268.7699890136719], ['2024-04-09', 265.6199951171875], ['2024-04-15', 260.4700012207031], ['2024-04-19', 252.05999755859375], ['2024-04-25', 258.5400085449219], ['2024-05-01', 259.7300109863281], ['2024-05-07', 268.3500061035156], ['2024-05-13', 270.0899963378906], ['2024-05-17', 259.1000061035156], ['2024-05-23', 257.2099914550781], ['2024-05-30', 254.0800018310547], ['2024-06-05', 251.17999267578125], ['2024-06-11', 244.63999938964844], ['2024-06-17', 256.3500061035156], ['2024-06-24', 247.5800018310547], ['2024-06-28', 248.67999267578125], ['2024-07-05', 244.2100067138672], ['2024-07-11', 254.69000244140625], ['2024-07-17', 255.8000030517578], ['2024-07-23', 260.1499938964844], ['2024-07-29', 264.5299987792969], ['2024-08-02', 266.3399963378906], ['2024-08-08', 246.3300018310547], ['2024-08-14', 242.5], ['2024-08-20', 245.60000610351562], ['2024-08-26', 250.67999267578125], ['2024-08-30', 245.2100067138672], ['2024-09-06', 231.97000122070312], ['2024-09-12', 236.27000427246094]]</details></pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 1900 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "You have access to the following tools: [{'type': 'function', 'function': {'name': 'get_stock_ticker', 'description': 'Get company stock ticker from its name.', 'parameters': {'properties': {'company_name': {'type': 'string'}}, 'required': ['company_name'], 'type': 'object'}}}, {'type': 'function', 'function': {'name': 'get_stock_data', 'description': \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\", 'parameters': {'properties': {'symbol': {'type': 'string'}, 'start_date': {'type': 'string'}, 'end_date': {'type': 'string'}}, 'required': ['symbol', 'start_date', 'end_date'], 'type': 'object'}}}]\n",
       "--\n",
       "user: Tell me about Vulcan Materials in 3 sentences\n",
       "--\n",
       "assistant: Here's a plan on how to use the provided tools to help the user learn about Vulcan Materials' financials:\n",
       "\n",
       "1. First, use the `get_stock_ticker` function to retrieve the stock ticker symbol for Vulcan Materials by passing \"Vulcan Materials\" as the `company_name` parameter.\n",
       "\n",
       "2. Once the stock ticker symbol is obtained, use the `get_stock_data` function to retrieve the historical stock prices for Vulcan Materials over a specified date range, such as the past year or past quarter.\n",
       "\n",
       "3. Analyze the retrieved stock price data to identify trends, patterns, and key events that may have impacted the company's financial performance.\n",
       "\n",
       "4. Provide the user with a summary of the company's financial performance, including its current stock price, historical price trends, and any notable events or announcements that may have affected its stock price.\n",
       "\n",
       "5. Offer additional insights or recommendations based on the analyzed data, such as whether the company's stock appears to be undervalued or overvalued, or whether there are any potential risks or opportunities that investors should be aware of.\n",
       "\n",
       "6. If the user requests more information, use the `get_stock_data` function to retrieve additional data, such as the company's dividend history, earnings per share, or other relevant financial metrics.\n",
       "\n",
       "7. Provide the user with visualizations or charts to help illustrate the company's financial performance and trends, making it easier to understand and interpret the data.\n",
       "\n",
       "By following this plan, we can leverage the provided tools to provide the user with a comprehensive overview of Vulcan Materials' financials and help them make more informed investment decisions.\n",
       "--\n",
       "assistant: {\"kind\": \"tool_call\", \"tool_name\": \"get_stock_ticker\", \"parameters\": {\"company_name\": \"Vulcan Materials\"}}\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "assistant: {\"kind\": \"tool_call\", \"tool_name\": \"get_stock_data\", \"parameters\": {\"symbol\": \"VMC\", \"start_date\": \"2023-09-17\", \"end_date\": \"2024-09-17\"}}\n",
       "--\n",
       "tool: [['2023-09-18', 211.69000244140625], ['2023-09-22', 200.6199951171875], ['2023-09-28', 205.02999877929688], ['2023-10-04', 205.02999877929688], ['2023-10-10', 211.0], ['2023-10-16', 213.0399932861328], ['2023-10-20', 201.5500030517578], ['2023-10-26', 193.97000122070312], ['2023-11-01', 203.8000030517578], ['2023-11-07', 207.2100067138672], ['2023-11-13', 210.49000549316406], ['2023-11-17', 212.35000610351562], ['2023-11-24', 211.67999267578125], ['2023-11-30', 213.55999755859375], ['2023-12-06', 211.92999267578125], ['2023-12-12', 220.89999389648438], ['2023-12-18', 222.6999969482422], ['2023-12-22', 224.92999267578125], ['2023-12-29', 227.00999450683594], ['2024-01-05', 221.6199951171875], ['2024-01-11', 224.36000061035156], ['2024-01-18', 225.1199951171875], ['2024-01-24', 219.8000030517578], ['2024-01-30', 231.0500030517578], ['2024-02-05', 229.64999389648438], ['2024-02-09', 240.0], ['2024-02-15', 242.4600067138672], ['2024-02-22', 256.94000244140625], ['2024-02-28', 262.29998779296875], ['2024-03-05', 264.9800109863281], ['2024-03-11', 264.95001220703125], ['2024-03-15', 266.8599853515625], ['2024-03-21', 275.5899963378906], ['2024-03-27', 272.7900085449219], ['2024-04-03', 268.7699890136719], ['2024-04-09', 265.6199951171875], ['2024-04-15', 260.4700012207031], ['2024-04-19', 252.05999755859375], ['2024-04-25', 258.5400085449219], ['2024-05-01', 259.7300109863281], ['2024-05-07', 268.3500061035156], ['2024-05-13', 270.0899963378906], ['2024-05-17', 259.1000061035156], ['2024-05-23', 257.2099914550781], ['2024-05-30', 254.0800018310547], ['2024-06-05', 251.17999267578125], ['2024-06-11', 244.63999938964844], ['2024-06-17', 256.3500061035156], ['2024-06-24', 247.5800018310547], ['2024-06-28', 248.67999267578125], ['2024-07-05', 244.2100067138672], ['2024-07-11', 254.69000244140625], ['2024-07-17', 255.8000030517578], ['2024-07-23', 260.1499938964844], ['2024-07-29', 264.5299987792969], ['2024-08-02', 266.3399963378906], ['2024-08-08', 246.3300018310547], ['2024-08-14', 242.5], ['2024-08-20', 245.60000610351562], ['2024-08-26', 250.67999267578125], ['2024-08-30', 245.2100067138672], ['2024-09-06', 231.97000122070312], ['2024-09-12', 236.27000427246094]]\n",
       "--\n",
       "user: \n",
       "Follow the plan you created earlier. When you are done, respond to the user.\n",
       "If you want to call a or several tools, output JSON like this\n",
       "{\"kind\": \"tool_call\", \"tool_name\": \"...\", \"parameters\": \"... unquoted parameters json ...\"}\n",
       "If you have called all the tools in the plan, respond to the user with the JSON of the form\n",
       "{\"kind\": \"response\", \"content\": \"... you response ... \"}.\n",
       "Output ONE JSON OBJECT ONLY PER LINE ONLY AND NOTHING ELSE.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 145 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='{\"kind\": \"response\", \"content\": \"Vulcan Materials (VMC) is a leading producer of construction materials, including aggregates, asphalt, and ready-mixed concrete. As of September 17, 2024, the company\\'s stock is trading at around 236.27, which is a slight decrease from its recent high. Over the past year, VMC\\'s stock has experienced some volatility, with a high of 275.59 and a low of 193.97. Despite this, the company\\'s financial performance has remained strong, with revenue growth and stable profit margins. Overall, Vulcan Materials appears to be a solid investment opportunity for those looking to invest in the construction materials industry.\"}', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[8] Assistant</h4><pre class='step-text'>kind: assistant\n",
       "\n",
       "Vulcan Materials (VMC) is a leading producer of construction materials, including aggregates, asphalt, and ready-mixed concrete. As of September 17, 2024, the company's stock is trading at around 236.27, which is a slight decrease from its recent high. Over the past year, VMC's stock has experienced some volatility, with a high of 275.59 and a low of 193.97. Despite this, the company's financial performance has remained strong, with revenue growth and stable profit margins. Overall, Vulcan Materials appears to be a solid investment opportunity for those looking to invest in the construction materials industry.</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[9] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: plan\n",
       "</pre></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "final_tape2 = None\n",
    "print(\"Run LLAMA agent!\")\n",
    "for event in main_loop(\n",
    "    agent2, DialogTape() + [UserStep(content=\"Tell me about Vulcan Materials in 3 sentences\")], env, max_loops=3\n",
    "):\n",
    "    if ae := event.agent_event:\n",
    "        if ae.step:\n",
    "            print(ae.step.model_dump_json(indent=2))\n",
    "        if ae.final_tape:\n",
    "            final_tape2 = ae.final_tape\n",
    "    if event.observation:\n",
    "        print(event.observation.model_dump_json(indent=2))\n",
    "assert final_tape2 is not None\n",
    "print(\"Final tape:\")\n",
    "HTML(render_tape_with_prompts(final_tape2, PrettyRenderer()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Sometimes the above agent works well, but quite likely you are seeing that the LLAMA-based agent is having trouble. Let's try to help it. For reproducibility, we'll use a pre-recorded failed tape."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:11:12.988101Z",
     "iopub.status.busy": "2025-02-24T13:11:12.987901Z",
     "iopub.status.idle": "2025-02-24T13:11:25.001980Z",
     "shell.execute_reply": "2025-02-24T13:11:25.001302Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>.basic-renderer-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; white-space: pre-wrap; color: black;}.basic-prompt-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; color: black;}.episode-row { display: flex; align-items: end; }.agent-column { width: 50%; }.user-column { width: 25%; }.annotator-column { width: 25%; }.inner-tape-container { display: flex }.inner-tape-indent { width: 10%; }.inner-tape { width: 90%; }.agent_node { text-align: center; position: relative; margin: 12px 0; }.agent_node hr { border: 1px solid #e1e1e1; margin: 0 !important;}.agent_node span { position: absolute; right: 0; top: -0.6em; background: white; color: grey !important; padding: 0 10px; }</style><style>.observation { background-color: #baffc9;; }.error_observation { background-color: #dd0000; }.action { background-color: #cccccc; }.thought { background-color: #ffffdb; }.call { background-color: #ffffff; }.respond { background-color: #ffffff; }.step-header { margin: 2pt 2pt 2pt 0 !important; }.step-text { font-size: 12px; white-space: pre-wrap; word-wrap: break-word;}</style><h3 style='margin: 2px'>Metadata</h3><details> <summary>id: 4778ec8a-442a-4282-9c42-789def38b0b5</summary> <div class='basic-renderer-box'>author: Agent\n",
       "author_tape_id: null\n",
       "error: null\n",
       "id: 4778ec8a-442a-4282-9c42-789def38b0b5\n",
       "n_added_steps: 2\n",
       "parent_id: f6248f2d-71e3-4984-8630-8a1ff5528adf\n",
       "result: {}\n",
       "</div> </details><h3 style='margin: 2px'>Steps</h3><div class='basic-renderer-box observation'><h4 class='step-header'>[0] User</h4><pre class='step-text'>kind: user\n",
       "\n",
       "Tell me about Vulcan Materials in 3 sentences</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[1] Thought: AssistantThought</h4><pre class='step-text'>kind: assistant_thought\n",
       "\n",
       "<details><summary>1206 characters ...</summary>Here's a plan to help the user learn about Vulcan Materials using the available tools:\n",
       "\n",
       "1. **Get the stock ticker symbol**: Use the `get_stock_ticker` function to retrieve the stock ticker symbol for Vulcan Materials by passing 'Vulcan Materials' as the `company_name` parameter. This will give us the ticker symbol, which we'll use for further research.\n",
       "\n",
       "2. **Get historical stock data**: Use the `get_stock_data` function to retrieve the historical stock prices for Vulcan Materials. We'll pass the obtained ticker symbol, a start date (e.g., '2020-01-01'), and an end date (e.g., '2024-09-17') as parameters to get the stock prices over the specified period.\n",
       "\n",
       "3. **Analyze the stock data**: Examine the obtained stock prices to identify trends, patterns, and correlations. We can calculate key metrics such as the average stock price, highest and lowest prices, and the overall stock performance over the specified period.\n",
       "\n",
       "4. **Provide an overview of Vulcan Materials**: Combine the obtained data with additional information about Vulcan Materials, such as its industry, sector, market capitalization, and major competitors, to provide a comprehensive overview of the company.\n",
       "\n",
       "Let's execute this plan!</details></pre></div><div class='basic-renderer-box action'><h4 class='step-header'>[2] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_ticker\n",
       "    arguments: '{\"company_name\": \"Vulcan Materials\"}'\n",
       "  id: tool_call_0_node_starts_at_2\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[3] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[4] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: tool_call_0_node_starts_at_2\n",
       "\n",
       "VMC</pre></div><div class='basic-renderer-box action'><h4 class='step-header'>[5] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_data\n",
       "    arguments: '{\"symbol\": \"VMC\", \"start_date\": \"2020-01-01\", \"end_date\": \"2024-09-17\"}'\n",
       "  id: tool_call_0_node_starts_at_5\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[6] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[7] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: tool_call_0_node_starts_at_5\n",
       "\n",
       "<details><summary>2615 characters ...</summary>[('2020-01-02', 142.72000122070312), ('2020-01-27', 138.8800048828125), ('2020-02-19', 135.39999389648438), ('2020-03-12', 107.12000274658203), ('2020-04-03', 99.20999908447266), ('2020-04-28', 110.7699966430664), ('2020-05-20', 101.47000122070312), ('2020-06-12', 112.08999633789062), ('2020-07-07', 122.66999816894531), ('2020-07-29', 121.08000183105469), ('2020-08-20', 126.11000061035156), ('2020-09-14', 130.33999633789062), ('2020-10-06', 141.30999755859375), ('2020-10-28', 137.47999572753906), ('2020-11-19', 140.11000061035156), ('2020-12-14', 135.44000244140625), ('2021-01-07', 162.00999450683594), ('2021-02-01', 152.0), ('2021-02-24', 174.14999389648438), ('2021-03-18', 168.0500030517578), ('2021-04-12', 174.35000610351562), ('2021-05-04', 189.2100067138672), ('2021-05-26', 183.92999267578125), ('2021-06-18', 165.83999633789062), ('2021-07-13', 173.52999877929688), ('2021-08-04', 179.66000366210938), ('2021-08-26', 187.52999877929688), ('2021-09-20', 169.4199981689453), ('2021-10-12', 170.8300018310547), ('2021-11-03', 195.0), ('2021-11-26', 197.4499969482422), ('2021-12-20', 198.77999877929688), ('2022-01-12', 202.8800048828125), ('2022-02-04', 184.0399932861328), ('2022-03-01', 174.44000244140625), ('2022-03-23', 177.33999633789062), ('2022-04-14', 174.77000427246094), ('2022-05-09', 162.9499969482422), ('2022-06-01', 164.5399932861328), ('2022-06-24', 145.82000732421875), ('2022-07-19', 152.57000732421875), ('2022-08-10', 174.52000427246094), ('2022-09-01', 166.22000122070312), ('2022-09-26', 153.9499969482422), ('2022-10-18', 158.77999877929688), ('2022-11-09', 168.94000244140625), ('2022-12-02', 184.49000549316406), ('2022-12-27', 175.67999267578125), ('2023-01-20', 178.88999938964844), ('2023-02-13', 186.05999755859375), ('2023-03-08', 178.9199981689453), ('2023-03-30', 170.13999938964844), ('2023-04-24', 171.41000366210938), ('2023-05-16', 195.00999450683594), ('2023-06-08', 206.17999267578125), ('2023-07-03', 223.33999633789062), ('2023-07-26', 223.4600067138672), ('2023-08-17', 213.3300018310547), ('2023-09-11', 216.58999633789062), ('2023-10-03', 199.99000549316406), ('2023-10-25', 203.3699951171875), ('2023-11-16', 212.27000427246094), ('2023-12-11', 218.86000061035156), ('2024-01-04', 220.67999267578125), ('2024-01-29', 229.3699951171875), ('2024-02-21', 253.42999267578125), ('2024-03-14', 270.7300109863281), ('2024-04-08', 269.7200012207031), ('2024-04-30', 257.6300048828125), ('2024-05-22', 260.8699951171875), ('2024-06-14', 252.63999938964844), ('2024-07-10', 247.72000122070312), ('2024-08-01', 271.1300048828125), ('2024-08-23', 254.88999938964844)]</details></pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 2046 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "You have access to the following tools: [{'type': 'function', 'function': {'name': 'get_stock_ticker', 'description': 'Get company stock ticker from its name.', 'parameters': {'properties': {'company_name': {'type': 'string'}}, 'required': ['company_name'], 'type': 'object'}}}, {'type': 'function', 'function': {'name': 'get_stock_data', 'description': \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\", 'parameters': {'properties': {'symbol': {'type': 'string'}, 'start_date': {'type': 'string'}, 'end_date': {'type': 'string'}}, 'required': ['symbol', 'start_date', 'end_date'], 'type': 'object'}}}]\n",
       "--\n",
       "user: Tell me about Vulcan Materials in 3 sentences\n",
       "--\n",
       "assistant: Here's a plan to help the user learn about Vulcan Materials using the available tools:\n",
       "\n",
       "1. **Get the stock ticker symbol**: Use the `get_stock_ticker` function to retrieve the stock ticker symbol for Vulcan Materials by passing 'Vulcan Materials' as the `company_name` parameter. This will give us the ticker symbol, which we'll use for further research.\n",
       "\n",
       "2. **Get historical stock data**: Use the `get_stock_data` function to retrieve the historical stock prices for Vulcan Materials. We'll pass the obtained ticker symbol, a start date (e.g., '2020-01-01'), and an end date (e.g., '2024-09-17') as parameters to get the stock prices over the specified period.\n",
       "\n",
       "3. **Analyze the stock data**: Examine the obtained stock prices to identify trends, patterns, and correlations. We can calculate key metrics such as the average stock price, highest and lowest prices, and the overall stock performance over the specified period.\n",
       "\n",
       "4. **Provide an overview of Vulcan Materials**: Combine the obtained data with additional information about Vulcan Materials, such as its industry, sector, market capitalization, and major competitors, to provide a comprehensive overview of the company.\n",
       "\n",
       "Let's execute this plan!\n",
       "--\n",
       "assistant: {\"kind\": \"tool_call\", \"tool_name\": \"get_stock_ticker\", \"parameters\": {\"company_name\": \"Vulcan Materials\"}}\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "assistant: {\"kind\": \"tool_call\", \"tool_name\": \"get_stock_data\", \"parameters\": {\"symbol\": \"VMC\", \"start_date\": \"2020-01-01\", \"end_date\": \"2024-09-17\"}}\n",
       "--\n",
       "tool: [('2020-01-02', 142.72000122070312), ('2020-01-27', 138.8800048828125), ('2020-02-19', 135.39999389648438), ('2020-03-12', 107.12000274658203), ('2020-04-03', 99.20999908447266), ('2020-04-28', 110.7699966430664), ('2020-05-20', 101.47000122070312), ('2020-06-12', 112.08999633789062), ('2020-07-07', 122.66999816894531), ('2020-07-29', 121.08000183105469), ('2020-08-20', 126.11000061035156), ('2020-09-14', 130.33999633789062), ('2020-10-06', 141.30999755859375), ('2020-10-28', 137.47999572753906), ('2020-11-19', 140.11000061035156), ('2020-12-14', 135.44000244140625), ('2021-01-07', 162.00999450683594), ('2021-02-01', 152.0), ('2021-02-24', 174.14999389648438), ('2021-03-18', 168.0500030517578), ('2021-04-12', 174.35000610351562), ('2021-05-04', 189.2100067138672), ('2021-05-26', 183.92999267578125), ('2021-06-18', 165.83999633789062), ('2021-07-13', 173.52999877929688), ('2021-08-04', 179.66000366210938), ('2021-08-26', 187.52999877929688), ('2021-09-20', 169.4199981689453), ('2021-10-12', 170.8300018310547), ('2021-11-03', 195.0), ('2021-11-26', 197.4499969482422), ('2021-12-20', 198.77999877929688), ('2022-01-12', 202.8800048828125), ('2022-02-04', 184.0399932861328), ('2022-03-01', 174.44000244140625), ('2022-03-23', 177.33999633789062), ('2022-04-14', 174.77000427246094), ('2022-05-09', 162.9499969482422), ('2022-06-01', 164.5399932861328), ('2022-06-24', 145.82000732421875), ('2022-07-19', 152.57000732421875), ('2022-08-10', 174.52000427246094), ('2022-09-01', 166.22000122070312), ('2022-09-26', 153.9499969482422), ('2022-10-18', 158.77999877929688), ('2022-11-09', 168.94000244140625), ('2022-12-02', 184.49000549316406), ('2022-12-27', 175.67999267578125), ('2023-01-20', 178.88999938964844), ('2023-02-13', 186.05999755859375), ('2023-03-08', 178.9199981689453), ('2023-03-30', 170.13999938964844), ('2023-04-24', 171.41000366210938), ('2023-05-16', 195.00999450683594), ('2023-06-08', 206.17999267578125), ('2023-07-03', 223.33999633789062), ('2023-07-26', 223.4600067138672), ('2023-08-17', 213.3300018310547), ('2023-09-11', 216.58999633789062), ('2023-10-03', 199.99000549316406), ('2023-10-25', 203.3699951171875), ('2023-11-16', 212.27000427246094), ('2023-12-11', 218.86000061035156), ('2024-01-04', 220.67999267578125), ('2024-01-29', 229.3699951171875), ('2024-02-21', 253.42999267578125), ('2024-03-14', 270.7300109863281), ('2024-04-08', 269.7200012207031), ('2024-04-30', 257.6300048828125), ('2024-05-22', 260.8699951171875), ('2024-06-14', 252.63999938964844), ('2024-07-10', 247.72000122070312), ('2024-08-01', 271.1300048828125), ('2024-08-23', 254.88999938964844)]\n",
       "--\n",
       "user: \n",
       "Follow the plan you created earlier. When you are done, respond to the user.\n",
       "If you want to call a or several tools, output JSON like this\n",
       "{\"kind\": \"tool_call\", \"tool_name\": \"...\", \"parameters\": \"... unquoted parameters json ...\"}\n",
       "If you have called all the tools in the plan, respond to the user with the JSON of the form\n",
       "{\"kind\": \"response\", \"content\": \"... you response ... \"}.\n",
       "Output ONE JSON OBJECT ONLY PER LINE ONLY AND NOTHING ELSE.\n",
       "REMEMBER: check what tool calls you have already made. Do not do the same call again!</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 155 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='{\"kind\": \"response\", \"content\": \"Vulcan Materials is a company that produces construction materials, including aggregates, asphalt, and concrete. The stock ticker symbol for Vulcan Materials is VMC. Based on the historical stock data from 2020-01-01 to 2024-09-17, the average stock price is around 174.48, with the highest price being 270.7300109863281 on 2024-03-14 and the lowest price being 99.20999908447266 on 2020-04-03. The company\\'s stock has shown significant growth over the past few years, with some fluctuations. It is essential to do thorough research and consider multiple factors before making any investment decisions.\"}', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[8] Assistant</h4><pre class='step-text'>kind: assistant\n",
       "\n",
       "Vulcan Materials is a company that produces construction materials, including aggregates, asphalt, and concrete. The stock ticker symbol for Vulcan Materials is VMC. Based on the historical stock data from 2020-01-01 to 2024-09-17, the average stock price is around 174.48, with the highest price being 270.7300109863281 on 2024-03-14 and the lowest price being 99.20999908447266 on 2020-04-03. The company's stock has shown significant growth over the past few years, with some fluctuations. It is essential to do thorough research and consider multiple factors before making any investment decisions.</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[9] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: plan\n",
       "</pre></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with open(\"assets/failed_tape.json\") as src:\n",
    "    failed_tape = DialogTape.model_validate(json.load(src))\n",
    "agent2b = agent2.model_copy(deep=True)\n",
    "agent2b.templates[\"call_or_respond\"] += (\n",
    "    \"REMEMBER: check what tool calls you have already made. Do not do the same call again!\"\n",
    ")\n",
    "resume_from_step8 = agent2b.run(failed_tape[:8]).get_final_tape()\n",
    "HTML(render_tape_with_prompts(resume_from_step8, PrettyRenderer()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We found that this helpful hint often gets LLAMA-based agent unstuck. Note how easy it was to test it thanks to the ability of the agent to resume from step 8!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4. Tape reuse and training data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "Another way to help this agent (or one with an even smaller LLM) is to finetune the LLM. And the most important step towards finetuning is making the training data!\n",
    "\n",
    "There are two ways to make training data in TapeAgents:\n",
    "- the basic one: use the `LLMCall` structures that the agent created when it produced the tape. You can retrieve them from the SQLite storage and convert into training text.\n",
    "- the much powerful one: call `agent.reuse` to reconstructed the prompts and outputs **and** to validate that with the reconstructed LLMCalls the agent would indeed create the given tape\n",
    "\n",
    "The big advantage of the second approach is that it allows you to use the tape from a teacher agent (think slower and more expensive) to train a student agent (think faster and cheaper). Or to train an agent on its own revised tapes.\n",
    "\n",
    "Of course, restrictions apply: the tape by agent A may not be reusable by agent B directly. You might have to add/remove some steps. But at least `agent.reuse` verifies if your tape modifications led to creation of a tape that the agent B can indeed produce."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Make training data from the past LLM calls"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:11:25.004754Z",
     "iopub.status.busy": "2025-02-24T13:11:25.004535Z",
     "iopub.status.idle": "2025-02-24T13:11:25.009903Z",
     "shell.execute_reply": "2025-02-24T13:11:25.009458Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Retrieved 4 LLM calls from the tape.\n",
      "From the first retrieved LLM call, the LLM will be trained to predict this text:\n",
      "---\n",
      "{\"kind\": \"tool_call\", \"tool_name\": \"get_stock_data\", \"parameters\": {\"symbol\": \"VMC\", \"start_date\": \"2023-09-17\", \"end_date\": \"2024-09-17\"}}<|eot_id|>\n"
     ]
    }
   ],
   "source": [
    "from tapeagents.observe import retrieve_tape_llm_calls\n",
    "\n",
    "llm_calls = retrieve_tape_llm_calls(final_tape2)\n",
    "print(f\"Retrieved {len(llm_calls)} LLM calls from the tape.\")\n",
    "# under the hood agent2 will route this request to its llm\n",
    "example_text = agent2.make_training_text(list(llm_calls.values())[0])\n",
    "print(\"From the first retrieved LLM call, the LLM will be trained to predict this text:\")\n",
    "print(\"---\")\n",
    "print(example_text.output_text)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Make training data by reusing a tape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Note how `agent2` reuses the tape by `agent1`, even though they have very different prompt and output formats! \n",
    "\n",
    "You can inspect the reused tape below and see that the steps are the same as before, but the prompts and outputs are different."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:11:25.011994Z",
     "iopub.status.busy": "2025-02-24T13:11:25.011775Z",
     "iopub.status.idle": "2025-02-24T13:11:25.029591Z",
     "shell.execute_reply": "2025-02-24T13:11:25.029111Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>.basic-renderer-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; white-space: pre-wrap; color: black;}.basic-prompt-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; color: black;}.episode-row { display: flex; align-items: end; }.agent-column { width: 50%; }.user-column { width: 25%; }.annotator-column { width: 25%; }.inner-tape-container { display: flex }.inner-tape-indent { width: 10%; }.inner-tape { width: 90%; }.agent_node { text-align: center; position: relative; margin: 12px 0; }.agent_node hr { border: 1px solid #e1e1e1; margin: 0 !important;}.agent_node span { position: absolute; right: 0; top: -0.6em; background: white; color: grey !important; padding: 0 10px; }</style><style>.observation { background-color: #baffc9;; }.error_observation { background-color: #dd0000; }.action { background-color: #cccccc; }.thought { background-color: #ffffdb; }.call { background-color: #ffffff; }.respond { background-color: #ffffff; }.step-header { margin: 2pt 2pt 2pt 0 !important; }.step-text { font-size: 12px; white-space: pre-wrap; word-wrap: break-word;}</style><h3 style='margin: 2px'>Metadata</h3><details> <summary>id: b1672333-e54f-4b72-b393-b18228615e20</summary> <div class='basic-renderer-box'>author: null\n",
       "author_tape_id: null\n",
       "error: null\n",
       "id: b1672333-e54f-4b72-b393-b18228615e20\n",
       "n_added_steps: 0\n",
       "parent_id: null\n",
       "result: {}\n",
       "</div> </details><h3 style='margin: 2px'>Steps</h3><div class='basic-renderer-box observation'><h4 class='step-header'>[0] User</h4><pre class='step-text'>kind: user\n",
       "\n",
       "Tell me about Vulcan in 3 sentences</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt -1 tokens , cached</summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "You have access to the following tools: [{'type': 'function', 'function': {'name': 'get_stock_ticker', 'description': 'Get company stock ticker from its name.', 'parameters': {'properties': {'company_name': {'type': 'string'}}, 'required': ['company_name'], 'type': 'object'}}}, {'type': 'function', 'function': {'name': 'get_stock_data', 'description': \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\", 'parameters': {'properties': {'symbol': {'type': 'string'}, 'start_date': {'type': 'string'}, 'end_date': {'type': 'string'}}, 'required': ['symbol', 'start_date', 'end_date'], 'type': 'object'}}}]\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "user: Write a natural language plan on how to use tools help the user. Output a list of numbered items, like 1., 2., 3., etc.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output -1 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\\n\\n2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\\n\\n3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\\n\\n4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box thought'><h4 class='step-header'>[1] Thought: AssistantThought</h4><pre class='step-text'>kind: assistant_thought\n",
       "\n",
       "1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\n",
       "\n",
       "2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\n",
       "\n",
       "3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\n",
       "\n",
       "4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt -1 tokens , cached</summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "You have access to the following tools: [{'type': 'function', 'function': {'name': 'get_stock_ticker', 'description': 'Get company stock ticker from its name.', 'parameters': {'properties': {'company_name': {'type': 'string'}}, 'required': ['company_name'], 'type': 'object'}}}, {'type': 'function', 'function': {'name': 'get_stock_data', 'description': \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\", 'parameters': {'properties': {'symbol': {'type': 'string'}, 'start_date': {'type': 'string'}, 'end_date': {'type': 'string'}}, 'required': ['symbol', 'start_date', 'end_date'], 'type': 'object'}}}]\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\n",
       "\n",
       "2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\n",
       "\n",
       "3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\n",
       "\n",
       "4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.\n",
       "--\n",
       "user: \n",
       "Follow the plan you created earlier. When you are done, respond to the user.\n",
       "If you want to call a or several tools, output JSON like this\n",
       "{\"kind\": \"tool_call\", \"tool_name\": \"...\", \"parameters\": \"... unquoted parameters json ...\"}\n",
       "If you have called all the tools in the plan, respond to the user with the JSON of the form\n",
       "{\"kind\": \"response\", \"content\": \"... you response ... \"}.\n",
       "Output ONE JSON OBJECT ONLY PER LINE ONLY AND NOTHING ELSE.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output -1 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='{\"kind\": \"tool_call\", \"tool_name\": \"get_stock_ticker\", \"parameters\": {\"company_name\": \"Vulcan\"}}', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[2] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_ticker\n",
       "    arguments: '{\"company_name\": \"Vulcan\"}'\n",
       "  id: tool_call_0_node_starts_at_2\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[3] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[4] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_al69pbS0aE2LrEXiYTarQ2ZJ\n",
       "\n",
       "VMC</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt -1 tokens , cached</summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "You have access to the following tools: [{'type': 'function', 'function': {'name': 'get_stock_ticker', 'description': 'Get company stock ticker from its name.', 'parameters': {'properties': {'company_name': {'type': 'string'}}, 'required': ['company_name'], 'type': 'object'}}}, {'type': 'function', 'function': {'name': 'get_stock_data', 'description': \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\", 'parameters': {'properties': {'symbol': {'type': 'string'}, 'start_date': {'type': 'string'}, 'end_date': {'type': 'string'}}, 'required': ['symbol', 'start_date', 'end_date'], 'type': 'object'}}}]\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\n",
       "\n",
       "2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\n",
       "\n",
       "3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\n",
       "\n",
       "4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.\n",
       "--\n",
       "assistant: {\"kind\": \"tool_call\", \"tool_name\": \"get_stock_ticker\", \"parameters\": {\"company_name\": \"Vulcan\"}}\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "user: \n",
       "Follow the plan you created earlier. When you are done, respond to the user.\n",
       "If you want to call a or several tools, output JSON like this\n",
       "{\"kind\": \"tool_call\", \"tool_name\": \"...\", \"parameters\": \"... unquoted parameters json ...\"}\n",
       "If you have called all the tools in the plan, respond to the user with the JSON of the form\n",
       "{\"kind\": \"response\", \"content\": \"... you response ... \"}.\n",
       "Output ONE JSON OBJECT ONLY PER LINE ONLY AND NOTHING ELSE.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output -1 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='{\"kind\": \"tool_call\", \"tool_name\": \"get_stock_data\", \"parameters\": {\"symbol\": \"VMC\", \"start_date\": \"2024-09-01\", \"end_date\": \"2024-09-16\"}}', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[5] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_data\n",
       "    arguments: '{\"symbol\": \"VMC\", \"start_date\": \"2024-09-01\", \"end_date\": \"2024-09-16\"}'\n",
       "  id: tool_call_0_node_starts_at_5\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[6] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[7] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_s0VrNLHmJksS9Mgvh4tgmXCb\n",
       "\n",
       "[['2024-09-03', 239.02000427246094], ['2024-09-04', 234.8300018310547], ['2024-09-05', 232.16000366210938], ['2024-09-06', 231.97000122070312], ['2024-09-09', 231.8300018310547], ['2024-09-10', 232.9499969482422], ['2024-09-11', 232.9499969482422], ['2024-09-12', 236.27000427246094], ['2024-09-13', 237.47000122070312]]</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt -1 tokens , cached</summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>system: \n",
       "You will help the user to learn about financials of companies.\n",
       "Use as many relevant tools as possible to include more details and facts in your responses.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "You have access to the following tools: [{'type': 'function', 'function': {'name': 'get_stock_ticker', 'description': 'Get company stock ticker from its name.', 'parameters': {'properties': {'company_name': {'type': 'string'}}, 'required': ['company_name'], 'type': 'object'}}}, {'type': 'function', 'function': {'name': 'get_stock_data', 'description': \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\", 'parameters': {'properties': {'symbol': {'type': 'string'}, 'start_date': {'type': 'string'}, 'end_date': {'type': 'string'}}, 'required': ['symbol', 'start_date', 'end_date'], 'type': 'object'}}}]\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\n",
       "\n",
       "2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\n",
       "\n",
       "3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\n",
       "\n",
       "4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.\n",
       "--\n",
       "assistant: {\"kind\": \"tool_call\", \"tool_name\": \"get_stock_ticker\", \"parameters\": {\"company_name\": \"Vulcan\"}}\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "assistant: {\"kind\": \"tool_call\", \"tool_name\": \"get_stock_data\", \"parameters\": {\"symbol\": \"VMC\", \"start_date\": \"2024-09-01\", \"end_date\": \"2024-09-16\"}}\n",
       "--\n",
       "tool: [['2024-09-03', 239.02000427246094], ['2024-09-04', 234.8300018310547], ['2024-09-05', 232.16000366210938], ['2024-09-06', 231.97000122070312], ['2024-09-09', 231.8300018310547], ['2024-09-10', 232.9499969482422], ['2024-09-11', 232.9499969482422], ['2024-09-12', 236.27000427246094], ['2024-09-13', 237.47000122070312]]\n",
       "--\n",
       "user: \n",
       "Follow the plan you created earlier. When you are done, respond to the user.\n",
       "If you want to call a or several tools, output JSON like this\n",
       "{\"kind\": \"tool_call\", \"tool_name\": \"...\", \"parameters\": \"... unquoted parameters json ...\"}\n",
       "If you have called all the tools in the plan, respond to the user with the JSON of the form\n",
       "{\"kind\": \"response\", \"content\": \"... you response ... \"}.\n",
       "Output ONE JSON OBJECT ONLY PER LINE ONLY AND NOTHING ELSE.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output -1 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content='{\"kind\": \"response\", \"content\": \"Vulcan Materials Company, trading under the stock ticker symbol VMC, is a major producer of construction aggregates and a leading producer of other construction materials. Recently, its stock prices have shown some fluctuations, with prices ranging from approximately $231.83 to $239.02 between September 1 and September 13, 2024. This reflects a stable yet slightly volatile market performance, indicative of the construction industry\\'s dynamics and economic conditions.\"}', role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[8] Assistant</h4><pre class='step-text'>kind: assistant\n",
       "\n",
       "Vulcan Materials Company, trading under the stock ticker symbol VMC, is a major producer of construction aggregates and a leading producer of other construction materials. Recently, its stock prices have shown some fluctuations, with prices ranging from approximately $231.83 to $239.02 between September 1 and September 13, 2024. This reflects a stable yet slightly volatile market performance, indicative of the construction industry's dynamics and economic conditions.</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[9] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: plan\n",
       "</pre></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reused_tape, _ = agent2.reuse(final_tape1)\n",
    "HTML(render_tape_with_prompts(reused_tape, PrettyRenderer()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We offer a quick way to harness the tape reuse to make training data. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:11:25.031478Z",
     "iopub.status.busy": "2025-02-24T13:11:25.031313Z",
     "iopub.status.idle": "2025-02-24T13:11:25.041705Z",
     "shell.execute_reply": "2025-02-24T13:11:25.041268Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1. Identify that the user is asking about a company named \"Vulcan\" and wants information about it.\n",
      "\n",
      "2. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan, as this will help in retrieving financial data.\n",
      "\n",
      "3. Once the stock ticker is obtained, use the `functions.get_stock_data` tool to gather recent stock price data for Vulcan to provide insights into its financial performance.\n",
      "\n",
      "4. Compile the information gathered from the tools to provide a concise summary about Vulcan, including its stock performance and any other relevant financial details.<|eot_id|>\n"
     ]
    }
   ],
   "source": [
    "training_data = agent2.make_training_data(final_tape1)\n",
    "print(training_data[0].output_text)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "What could be simpler than that?!\n",
    "We're only scratching the surface of what TapeAgents can do. We invite you to explore the other tutorials and examples to learn more about the TapeAgents framework."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5. Multi-agent teams, deeper dive into view stack"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's add a colleague to our agent that will help it search the internet! \n",
    "\n",
    "First thing, we need to give this colleague some tools. In TapeAgents, the agents do not use the environment directly, their interaction with the environment is mediated by in orchestrator such as you application. But the agents should know what tools they can call. And if you are using `main_loop` as the orchestrator, it requires one environment that contains tools of all the agents. \n",
    "\n",
    "Let's go ahead and define environments:\n",
    "- the environments of the root agent\n",
    "- the one of its internet search specialist colleague\n",
    "- the environment that contains all the tools.\n",
    "\n",
    "Note that we won't use the first two environments to produce observations, we'll use them only to generate tool schemas for agent."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:11:25.043352Z",
     "iopub.status.busy": "2025-02-24T13:11:25.043203Z",
     "iopub.status.idle": "2025-02-24T13:11:25.397276Z",
     "shell.execute_reply": "2025-02-24T13:11:25.396905Z"
    }
   },
   "outputs": [],
   "source": [
    "from tapeagents.tools.simple_browser import SimpleTextBrowser\n",
    "from tapeagents.tools.web_search import web_search_tool\n",
    "\n",
    "browser = SimpleTextBrowser()\n",
    "search_agent_env = ToolEnvironment([web_search_tool, browser.get_page, browser.get_next_page])\n",
    "\n",
    "\n",
    "# We will use the tool choice mechanism to let the main agent call its search specialist agent.\n",
    "# To this end, we create a mock tool that represents calling the search agent.\n",
    "def call_search_agent(query: str):\n",
    "    \"\"\"Use this tool to ask a fellow AI agent to search for information on the web.\"\"\"\n",
    "    pass\n",
    "\n",
    "\n",
    "main_agent_env = ToolEnvironment([get_stock_ticker, get_stock_data, call_search_agent])\n",
    "whole_env = ToolEnvironment(\n",
    "    [get_stock_ticker, get_stock_data, web_search_tool, browser.get_page, browser.get_next_page]\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Before we implement the subagent, let's review the way we do multi-agent communication in TapeAgents:\n",
    "- when the root agent wants to call its subagent \"xyz\", it puts `Call(agent_name=\"xyz\")` step on the tape\n",
    "- at the next iteration, the root agent will compute `TapeViewStack` and delegate to the currently active agent. Right after `Call(agent_name=\"B\")` on top of the stack there will be a new view associated with Agent \"xyz\". The root agent will delegate to \"xyz\" to make the prompt and to generate the steps from \"xyz\"'s current node.\n",
    "- when \"xyz\" is done it will put `Respond()` step on the tape. At the next iteration the view stack won't have \"xyz\"'s view on the top any more. \n",
    "\n",
    "A reader familiar with the concept of a call stack will find the `TapeViewStack` concept very similar...\n",
    "\n",
    "Note that for the purpose of computing the view stack the root agent's name is not known. It's view will be signed as \"root\".\n",
    "\n",
    "Let's explore the way `Call` and `Respond` works with a minimal example.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:11:25.398796Z",
     "iopub.status.busy": "2025-02-24T13:11:25.398620Z",
     "iopub.status.idle": "2025-02-24T13:11:25.401574Z",
     "shell.execute_reply": "2025-02-24T13:11:25.401335Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "View stack after step 0\n",
      "-- root: UserStep\n",
      "View stack after step 1\n",
      "-- root: UserStep, Call\n",
      "-- root/xyz: Call\n",
      "View stack after step 2\n",
      "-- root: UserStep, Call\n",
      "-- root/xyz: Call, AssistantThought\n",
      "View stack after step 3\n",
      "-- root: UserStep, Call, Respond\n"
     ]
    }
   ],
   "source": [
    "from tapeagents.core import Call, Respond\n",
    "\n",
    "tape = DialogTape(\n",
    "    steps=[\n",
    "        UserStep(content=\"Compute 2 + 2\"),\n",
    "        Call(agent_name=\"xyz\", content=\"what is 2 + 2, dear xyz\"),\n",
    "        AssistantThought(content=\"deep thinking by agent xyz\"),\n",
    "        Respond(content=\"I heard it is 4\"),\n",
    "    ]\n",
    ")\n",
    "# We will print a brief summary of view stack after each step\n",
    "for i in range(0, len(tape)):\n",
    "    print(f\"View stack after step {i}\")\n",
    "    view_stack = TapeViewStack.compute(tape[: i + 1])\n",
    "    for view in view_stack.stack:\n",
    "        step_summary = \", \".join([step.__class__.__name__ for step in view.steps])\n",
    "        print(f\"-- {view.agent_full_name}: {step_summary}\")\n",
    "# Note how \"root/xyz\" view appears after step 1 and disappears after step 3.\n",
    "# Also note how \"root\" does not see private thoughts of \"xyz\" (step 2), and \"xyz\"\n",
    "# does not see the initial observation of root (step 0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We are now ready to proceed to the implementation of the internet search agent!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:11:25.402709Z",
     "iopub.status.busy": "2025-02-24T13:11:25.402634Z",
     "iopub.status.idle": "2025-02-24T13:11:39.306348Z",
     "shell.execute_reply": "2025-02-24T13:11:39.305917Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>.basic-renderer-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; white-space: pre-wrap; color: black;}.basic-prompt-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; color: black;}.episode-row { display: flex; align-items: end; }.agent-column { width: 50%; }.user-column { width: 25%; }.annotator-column { width: 25%; }.inner-tape-container { display: flex }.inner-tape-indent { width: 10%; }.inner-tape { width: 90%; }.agent_node { text-align: center; position: relative; margin: 12px 0; }.agent_node hr { border: 1px solid #e1e1e1; margin: 0 !important;}.agent_node span { position: absolute; right: 0; top: -0.6em; background: white; color: grey !important; padding: 0 10px; }</style><style>.observation { background-color: #baffc9;; }.error_observation { background-color: #dd0000; }.action { background-color: #cccccc; }.thought { background-color: #ffffdb; }.call { background-color: #ffffff; }.respond { background-color: #ffffff; }.step-header { margin: 2pt 2pt 2pt 0 !important; }.step-text { font-size: 12px; white-space: pre-wrap; word-wrap: break-word;}</style><h3 style='margin: 2px'>Metadata</h3><details> <summary>id: 2676ed10-0234-4bca-9b98-14c3d230efb1</summary> <div class='basic-renderer-box'>author: null\n",
       "author_tape_id: null\n",
       "error: null\n",
       "id: 2676ed10-0234-4bca-9b98-14c3d230efb1\n",
       "n_added_steps: 0\n",
       "parent_id: null\n",
       "result: {}\n",
       "</div> </details><h3 style='margin: 2px'>Steps</h3><div class='basic-renderer-box call'><h4 class='step-header'>[0] Agent calls search_agent</h4><pre class='step-text'>kind: call\n",
       "agent_name: search_agent\n",
       "\n",
       "What influenced Nvidia stock price in late 2022?</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 22 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"web_search_tool\",\n",
       "      \"description\": \"Search the web for a given query, return a list of search result dictionaries.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"max_results\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"retry_pause\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"attempts\": {\n",
       "            \"default\": 3,\n",
       "            \"type\": \"integer\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_page\",\n",
       "      \"description\": \"Loads a web page and returns a tuple containing:\\n- Content of its first viewport (first screen)\\n- Current page number\\n- Total number of pages\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"url\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"url\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_next_page\",\n",
       "      \"description\": \"Load next page of the document and return the current content of the viewport, current page number and total number of pages.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {},\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "user (Agent): What influenced Nvidia stock price in late 2022?</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 0 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=None, role='assistant', tool_calls=[ChatCompletionMessageToolCall(function=Function(arguments='{\"query\":\"Nvidia stock price influences late 2022\",\"max_results\":5}', name='web_search_tool'), id='call_0ck4l7FiYey1zBXJP8tpSoki', type='function')], function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[1] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: web_search_tool\n",
       "    arguments: '{\"query\":\"Nvidia stock price influences late 2022\",\"max_results\":5}'\n",
       "  id: call_0ck4l7FiYey1zBXJP8tpSoki\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[2] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: main\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[3] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_0ck4l7FiYey1zBXJP8tpSoki\n",
       "\n",
       "<details><summary>1858 characters ...</summary>[{'title': 'NVDA Stock Has Gained 853% Since 2022 Fiscal End, Primarily ...', 'url': 'https://www.trefis.com/data/companies/NVDA/no-login-required/YZaWTk8F/Why-NVIDIA-Stock-Moved-NVDA-Stock-Has-Gained-853-Since-2022-Fiscal-End-Primarily-Due-To-Favorable-Change-In-Revenues', 'content': 'Factors Driving NVIDIA (NVDA) Stock Price Change: [1] Price To Sales Ratio: NVDA Price To Sales Ratio (P/S) Has Increased 42% Since 2022 End ...'}, {'title': 'Should You Buy Nvidia Stock Hand Over Fist Before Feb. 26?', 'url': 'https://www.fool.com/investing/2025/02/23/should-you-buy-nvidia-stock-hand-over-fist-before/', 'content': \"... 2022, Nvidia has a great track record of exceeding Wall Street's earnings expectations. And its share price has often performed well afterward.\"}, {'title': \"The Rise and Fall of Nvidia's Stock Price: A macro and ... - LinkedIn\", 'url': 'https://www.linkedin.com/pulse/rise-fall-nvidias-stock-price-macro-microeconomic-diego-nohuf', 'content': 'The stock price of Nvidia has been affected by a number of factors in recent years, including the COVID-19 pandemic, low interest rates, and the ...'}, {'title': 'NVDA Stock Has Gained 849% Since 2022 Fiscal End, Primarily ...', 'url': 'https://www.trefis.com/data/companies/NVDA/no-login-required/YZaWTk8F/Why-NVIDIA-Stock-Moved-NVDA-Stock-Has-Gained-319-Since-2020-Primarily-Due-To-Favorable-Change-In-Revenues', 'content': 'Factors Driving NVIDIA (NVDA) Stock Price Change: [1] Price To Sales Ratio: NVDA Price To Sales Ratio (P/S) Has Increased 41% Since 2022 End. NVDA Price to ...'}, {'title': 'Stock Info - Historical Price Lookup - NVIDIA Corporation', 'url': 'https://investor.nvidia.com/stock-info/historical-price-lookup/default.aspx', 'content': \"Stock Quote: NASD: NVDA. Day's Open. 140.04. Volume. 227,352,400. Intraday High. 141.46. Intraday Low. 134.03. Closing Price.\"}]</details></pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 568 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"web_search_tool\",\n",
       "      \"description\": \"Search the web for a given query, return a list of search result dictionaries.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"max_results\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"retry_pause\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"attempts\": {\n",
       "            \"default\": 3,\n",
       "            \"type\": \"integer\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_page\",\n",
       "      \"description\": \"Loads a web page and returns a tuple containing:\\n- Content of its first viewport (first screen)\\n- Current page number\\n- Total number of pages\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"url\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"url\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_next_page\",\n",
       "      \"description\": \"Load next page of the document and return the current content of the viewport, current page number and total number of pages.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {},\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "user (Agent): What influenced Nvidia stock price in late 2022?\n",
       "--\n",
       "assistant: [{'function': {'name': 'web_search_tool', 'arguments': '{\"query\":\"Nvidia stock price influences late 2022\",\"max_results\":5}'}, 'id': 'call_0ck4l7FiYey1zBXJP8tpSoki', 'type': 'function'}]\n",
       "--\n",
       "tool: [{'title': 'NVDA Stock Has Gained 853% Since 2022 Fiscal End, Primarily ...', 'url': 'https://www.trefis.com/data/companies/NVDA/no-login-required/YZaWTk8F/Why-NVIDIA-Stock-Moved-NVDA-Stock-Has-Gained-853-Since-2022-Fiscal-End-Primarily-Due-To-Favorable-Change-In-Revenues', 'content': 'Factors Driving NVIDIA (NVDA) Stock Price Change: [1] Price To Sales Ratio: NVDA Price To Sales Ratio (P/S) Has Increased 42% Since 2022 End ...'}, {'title': 'Should You Buy Nvidia Stock Hand Over Fist Before Feb. 26?', 'url': 'https://www.fool.com/investing/2025/02/23/should-you-buy-nvidia-stock-hand-over-fist-before/', 'content': \"... 2022, Nvidia has a great track record of exceeding Wall Street's earnings expectations. And its share price has often performed well afterward.\"}, {'title': \"The Rise and Fall of Nvidia's Stock Price: A macro and ... - LinkedIn\", 'url': 'https://www.linkedin.com/pulse/rise-fall-nvidias-stock-price-macro-microeconomic-diego-nohuf', 'content': 'The stock price of Nvidia has been affected by a number of factors in recent years, including the COVID-19 pandemic, low interest rates, and the ...'}, {'title': 'NVDA Stock Has Gained 849% Since 2022 Fiscal End, Primarily ...', 'url': 'https://www.trefis.com/data/companies/NVDA/no-login-required/YZaWTk8F/Why-NVIDIA-Stock-Moved-NVDA-Stock-Has-Gained-319-Since-2020-Primarily-Due-To-Favorable-Change-In-Revenues', 'content': 'Factors Driving NVIDIA (NVDA) Stock Price Change: [1] Price To Sales Ratio: NVDA Price To Sales Ratio (P/S) Has Increased 41% Since 2022 End. NVDA Price to ...'}, {'title': 'Stock Info - Historical Price Lookup - NVIDIA Corporation', 'url': 'https://investor.nvidia.com/stock-info/historical-price-lookup/default.aspx', 'content': \"Stock Quote: NASD: NVDA. Day's Open. 140.04. Volume. 227,352,400. Intraday High. 141.46. Intraday Low. 134.03. Closing Price.\"}]</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 0 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=None, role='assistant', tool_calls=[ChatCompletionMessageToolCall(function=Function(arguments='{\"url\":\"https://www.trefis.com/data/companies/NVDA/no-login-required/YZaWTk8F/Why-NVIDIA-Stock-Moved-NVDA-Stock-Has-Gained-853-Since-2022-Fiscal-End-Primarily-Due-To-Favorable-Change-In-Revenues\"}', name='get_page'), id='call_i8wbMeHjQ0GO4yilpynPbaAe', type='function')], function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[4] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_page\n",
       "    arguments: '{\"url\":\"https://www.trefis.com/data/companies/NVDA/no-login-required/YZaWTk8F/Why-NVIDIA-Stock-Moved-NVDA-Stock-Has-Gained-853-Since-2022-Fiscal-End-Primarily-Due-To-Favorable-Change-In-Revenues\"}'\n",
       "  id: call_i8wbMeHjQ0GO4yilpynPbaAe\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[5] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: main\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[6] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_i8wbMeHjQ0GO4yilpynPbaAe\n",
       "\n",
       "<details><summary>3024 characters ...</summary>('Title: Why NVIDIA Stock Moved: NVDA Stock Has Gained 821% Since 2022 Fiscal End, Primarily Due To Favorable Change In Revenues | Trefis\\n=======================\\nLoading Dashboard ...\\n\\n[![Trefis]()](https://www.trefis.com/data/home?from=icon)\\n\\n* [All Data](https://www.trefis.com/data/home)\\n* [About](https://www.trefis.com/info/about-trefis/)\\n\\t+ [Our Story](https://www.trefis.com/info/about-trefis/)\\n\\t+ [Management Team](https://www.trefis.com/info/about-trefis/management-team/)\\n\\t+ [Investors and Advisors](https://www.trefis.com/info/about-trefis/investors-and-advisors/)\\n\\t+ [Careers](https://www.trefis.com/info/about-trefis/careers/)\\n\\t+ [Contact Us](https://www.trefis.com/info/about-trefis/contact-us/)\\n\\n* Try Trefis\\n* Sign In\\n\\n×\\n\\n* Try Trefis\\n* Sign In\\n\\n* [All Data](https://www.trefis.com/data/home)\\n* [About](https://www.trefis.com/info/about-trefis/)\\n\\t+ [Our Story](https://www.trefis.com/info/about-trefis/)\\n\\t+ [Management Team](https://www.trefis.com/info/about-trefis/management-team/)\\n\\t+ [Investors and Advisors](https://www.trefis.com/info/about-trefis/investors-and-advisors/)\\n\\t+ [Careers](https://www.trefis.com/info/about-trefis/careers/)\\n\\t+ [Contact Us](https://www.trefis.com/info/about-trefis/contact-us/)\\n\\n* Try Trefis\\n* Sign In\\n\\nThis site requires JavaScript to function properly.\\n For help enabling JavaScript [click here](http://www.google.com/support/bin/answer.py?answer=23852).\\n---------------------------------------------------------------------------------------------------------------------------------------------------------\\n\\n### Here are some sample widgets you could interact with from this dashboard:\\n\\n#### EXPLORE TREFIS\\n\\n* [Home](https://www.trefis.com/?from=link)\\n* [About Us](https://www.trefis.com/info/about-trefis/)\\n* [Blog](https://www.trefis.com/info/blog/)\\n* [Resources](https://www.trefis.com/info/resources/)\\n* [Contact Us](https://www.trefis.com/info/about-trefis/contact-us/)\\n* [Careers](https://www.trefis.com/info/about-trefis/careers/)\\n* [Terms of Use](https://www.trefis.com/info/terms)\\n\\n©Copyright 2019 Insight Guru Inc. All Rights Reserved.\\n\\nBy using the Site, you agree to be bound by our Terms of\\n Use. Financial market data powered by Quotemedia.com.\\n Consensus EPS estimates are from QuoteMedia and are updated every weekday.\\n All rights reserved.\\n\\n NYSE/AMEX data delayed 20 minutes. NASDAQ and other\\n data delayed 15 minutes unless indicated.\\n\\n![](/tracking/beacon/0.27069512381351457?modelVersion=1466021&activity=MODEL&type=VIEW)\\n\\nThis interactive model has a limit on the number of drivers that can be modified in a single scenario. When the limit is reached those drivers not yet modified become disabled for modification.\\n\\nThis interactive model has a minimum and/or maximum limit on the value for the current driver you are viewing.\\n\\nThis driver cannot be modified within this scenario. This scenario is linked to a parent scenario that has already modified this driver.', 1, 0)</details></pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 1471 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"web_search_tool\",\n",
       "      \"description\": \"Search the web for a given query, return a list of search result dictionaries.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"max_results\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"retry_pause\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"attempts\": {\n",
       "            \"default\": 3,\n",
       "            \"type\": \"integer\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_page\",\n",
       "      \"description\": \"Loads a web page and returns a tuple containing:\\n- Content of its first viewport (first screen)\\n- Current page number\\n- Total number of pages\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"url\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"url\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_next_page\",\n",
       "      \"description\": \"Load next page of the document and return the current content of the viewport, current page number and total number of pages.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {},\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "user (Agent): What influenced Nvidia stock price in late 2022?\n",
       "--\n",
       "assistant: [{'function': {'name': 'web_search_tool', 'arguments': '{\"query\":\"Nvidia stock price influences late 2022\",\"max_results\":5}'}, 'id': 'call_0ck4l7FiYey1zBXJP8tpSoki', 'type': 'function'}]\n",
       "--\n",
       "tool: [{'title': 'NVDA Stock Has Gained 853% Since 2022 Fiscal End, Primarily ...', 'url': 'https://www.trefis.com/data/companies/NVDA/no-login-required/YZaWTk8F/Why-NVIDIA-Stock-Moved-NVDA-Stock-Has-Gained-853-Since-2022-Fiscal-End-Primarily-Due-To-Favorable-Change-In-Revenues', 'content': 'Factors Driving NVIDIA (NVDA) Stock Price Change: [1] Price To Sales Ratio: NVDA Price To Sales Ratio (P/S) Has Increased 42% Since 2022 End ...'}, {'title': 'Should You Buy Nvidia Stock Hand Over Fist Before Feb. 26?', 'url': 'https://www.fool.com/investing/2025/02/23/should-you-buy-nvidia-stock-hand-over-fist-before/', 'content': \"... 2022, Nvidia has a great track record of exceeding Wall Street's earnings expectations. And its share price has often performed well afterward.\"}, {'title': \"The Rise and Fall of Nvidia's Stock Price: A macro and ... - LinkedIn\", 'url': 'https://www.linkedin.com/pulse/rise-fall-nvidias-stock-price-macro-microeconomic-diego-nohuf', 'content': 'The stock price of Nvidia has been affected by a number of factors in recent years, including the COVID-19 pandemic, low interest rates, and the ...'}, {'title': 'NVDA Stock Has Gained 849% Since 2022 Fiscal End, Primarily ...', 'url': 'https://www.trefis.com/data/companies/NVDA/no-login-required/YZaWTk8F/Why-NVIDIA-Stock-Moved-NVDA-Stock-Has-Gained-319-Since-2020-Primarily-Due-To-Favorable-Change-In-Revenues', 'content': 'Factors Driving NVIDIA (NVDA) Stock Price Change: [1] Price To Sales Ratio: NVDA Price To Sales Ratio (P/S) Has Increased 41% Since 2022 End. NVDA Price to ...'}, {'title': 'Stock Info - Historical Price Lookup - NVIDIA Corporation', 'url': 'https://investor.nvidia.com/stock-info/historical-price-lookup/default.aspx', 'content': \"Stock Quote: NASD: NVDA. Day's Open. 140.04. Volume. 227,352,400. Intraday High. 141.46. Intraday Low. 134.03. Closing Price.\"}]\n",
       "--\n",
       "assistant: [{'function': {'name': 'get_page', 'arguments': '{\"url\":\"https://www.trefis.com/data/companies/NVDA/no-login-required/YZaWTk8F/Why-NVIDIA-Stock-Moved-NVDA-Stock-Has-Gained-853-Since-2022-Fiscal-End-Primarily-Due-To-Favorable-Change-In-Revenues\"}'}, 'id': 'call_i8wbMeHjQ0GO4yilpynPbaAe', 'type': 'function'}]\n",
       "--\n",
       "tool: ('Title: Why NVIDIA Stock Moved: NVDA Stock Has Gained 821% Since 2022 Fiscal End, Primarily Due To Favorable Change In Revenues | Trefis\\n=======================\\nLoading Dashboard ...\\n\\n[![Trefis]()](https://www.trefis.com/data/home?from=icon)\\n\\n* [All Data](https://www.trefis.com/data/home)\\n* [About](https://www.trefis.com/info/about-trefis/)\\n\\t+ [Our Story](https://www.trefis.com/info/about-trefis/)\\n\\t+ [Management Team](https://www.trefis.com/info/about-trefis/management-team/)\\n\\t+ [Investors and Advisors](https://www.trefis.com/info/about-trefis/investors-and-advisors/)\\n\\t+ [Careers](https://www.trefis.com/info/about-trefis/careers/)\\n\\t+ [Contact Us](https://www.trefis.com/info/about-trefis/contact-us/)\\n\\n* Try Trefis\\n* Sign In\\n\\n×\\n\\n* Try Trefis\\n* Sign In\\n\\n* [All Data](https://www.trefis.com/data/home)\\n* [About](https://www.trefis.com/info/about-trefis/)\\n\\t+ [Our Story](https://www.trefis.com/info/about-trefis/)\\n\\t+ [Management Team](https://www.trefis.com/info/about-trefis/management-team/)\\n\\t+ [Investors and Advisors](https://www.trefis.com/info/about-trefis/investors-and-advisors/)\\n\\t+ [Careers](https://www.trefis.com/info/about-trefis/careers/)\\n\\t+ [Contact Us](https://www.trefis.com/info/about-trefis/contact-us/)\\n\\n* Try Trefis\\n* Sign In\\n\\nThis site requires JavaScript to function properly.\\n For help enabling JavaScript [click here](http://www.google.com/support/bin/answer.py?answer=23852).\\n---------------------------------------------------------------------------------------------------------------------------------------------------------\\n\\n### Here are some sample widgets you could interact with from this dashboard:\\n\\n#### EXPLORE TREFIS\\n\\n* [Home](https://www.trefis.com/?from=link)\\n* [About Us](https://www.trefis.com/info/about-trefis/)\\n* [Blog](https://www.trefis.com/info/blog/)\\n* [Resources](https://www.trefis.com/info/resources/)\\n* [Contact Us](https://www.trefis.com/info/about-trefis/contact-us/)\\n* [Careers](https://www.trefis.com/info/about-trefis/careers/)\\n* [Terms of Use](https://www.trefis.com/info/terms)\\n\\n©Copyright 2019 Insight Guru Inc. All Rights Reserved.\\n\\nBy using the Site, you agree to be bound by our Terms of\\n Use. Financial market data powered by Quotemedia.com.\\n Consensus EPS estimates are from QuoteMedia and are updated every weekday.\\n All rights reserved.\\n\\n NYSE/AMEX data delayed 20 minutes. NASDAQ and other\\n data delayed 15 minutes unless indicated.\\n\\n![](/tracking/beacon/0.27069512381351457?modelVersion=1466021&activity=MODEL&type=VIEW)\\n\\nThis interactive model has a limit on the number of drivers that can be modified in a single scenario. When the limit is reached those drivers not yet modified become disabled for modification.\\n\\nThis interactive model has a minimum and/or maximum limit on the value for the current driver you are viewing.\\n\\nThis driver cannot be modified within this scenario. This scenario is linked to a parent scenario that has already modified this driver.', 1, 0)</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 246 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=\"The search results and the page I accessed did not provide specific details about the factors influencing Nvidia's stock price in late 2022. However, based on general knowledge, several factors could have influenced Nvidia's stock price during that period:\\n\\n1. **Market Conditions**: The overall market conditions, including interest rates, inflation, and economic outlook, can significantly impact stock prices.\\n\\n2. **Company Performance**: Nvidia's financial performance, including revenue, profit margins, and earnings reports, would directly affect its stock price.\\n\\n3. **Product Releases and Innovations**: New product launches or technological advancements can boost investor confidence and drive stock prices up.\\n\\n4. **Supply Chain Issues**: Any disruptions in the supply chain, especially in the semiconductor industry, could impact Nvidia's production and sales, influencing its stock price.\\n\\n5. **Geopolitical Factors**: Trade tensions, especially between the US and China, could affect Nvidia's operations and stock price.\\n\\n6. **Investor Sentiment**: Changes in investor sentiment, driven by analyst ratings or market speculation, could also influence stock prices.\\n\\nFor more detailed and specific information, you might want to look at financial news articles or reports from late 2022 that focus on Nvidia.\", role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box respond'><h4 class='step-header'>[7] search_agent responds to Agent</h4><pre class='step-text'>kind: respond\n",
       "copy_output: false\n",
       "\n",
       "<details><summary>1319 characters ...</summary>The search results and the page I accessed did not provide specific details about the factors influencing Nvidia's stock price in late 2022. However, based on general knowledge, several factors could have influenced Nvidia's stock price during that period:\n",
       "\n",
       "1. **Market Conditions**: The overall market conditions, including interest rates, inflation, and economic outlook, can significantly impact stock prices.\n",
       "\n",
       "2. **Company Performance**: Nvidia's financial performance, including revenue, profit margins, and earnings reports, would directly affect its stock price.\n",
       "\n",
       "3. **Product Releases and Innovations**: New product launches or technological advancements can boost investor confidence and drive stock prices up.\n",
       "\n",
       "4. **Supply Chain Issues**: Any disruptions in the supply chain, especially in the semiconductor industry, could impact Nvidia's production and sales, influencing its stock price.\n",
       "\n",
       "5. **Geopolitical Factors**: Trade tensions, especially between the US and China, could affect Nvidia's operations and stock price.\n",
       "\n",
       "6. **Investor Sentiment**: Changes in investor sentiment, driven by analyst ratings or market speculation, could also influence stock prices.\n",
       "\n",
       "For more detailed and specific information, you might want to look at financial news articles or reports from late 2022 that focus on Nvidia.</details></pre></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from tapeagents.core import Respond\n",
    "from tapeagents.nodes import FixedStepsNode\n",
    "from tapeagents.prompting import view_to_messages\n",
    "\n",
    "search_system_instruction = \"Use at most 5 tool calls to search the request info on on the web.\"\n",
    "search_system_message = {\"role\": \"system\", \"content\": search_system_instruction}\n",
    "\n",
    "\n",
    "class SearchAgentMainNode(Node):\n",
    "    name: str = \"main\"\n",
    "\n",
    "    def make_prompt(self, agent, tape: DialogTape) -> Prompt:\n",
    "        view = agent.compute_view(tape)\n",
    "        return Prompt(messages=view_to_messages(view.top, agent), tools=search_agent_env.get_tool_schema_dicts())\n",
    "\n",
    "    def generate_steps(self, agent, tape, llm_stream: LLMStream):\n",
    "        o = llm_stream.get_output()\n",
    "        if o.content:\n",
    "            # if the LLM responds, yield Respond(..) as your last step\n",
    "            yield Respond(content=o.content)\n",
    "        elif o.tool_calls:\n",
    "            # when the LLM suggests tool calls, yield them as action steps\n",
    "            yield ToolCalls.from_llm_output(o)\n",
    "            yield SetNextNode(next_node=\"main\")\n",
    "        else:\n",
    "            raise ValueError()\n",
    "\n",
    "\n",
    "search_agent = Agent.create(\n",
    "    name=\"search_agent\",\n",
    "    llms=LiteLLM(model_name=\"gpt-4o\", parameters={\"temperature\": 0.1}),\n",
    "    nodes=[SearchAgentMainNode()],\n",
    ")\n",
    "# To test the subagent, we'll make a mock root agent that immediately calls \"search_agent\"\n",
    "call_step = Call(agent_name=\"search_agent\", content=\"What influenced Nvidia stock price in late 2022?\")\n",
    "test_root_agent = Agent.create(subagents=[search_agent], nodes=[FixedStepsNode(steps=[call_step])])\n",
    "start_tape = DialogTape()\n",
    "final_tape = None\n",
    "for event in main_loop(test_root_agent, start_tape, search_agent_env):\n",
    "    # We need to stop the loop when \"search_agent\" responds,\n",
    "    # otherwise the code will crash because `test_root_agent` only has one node.\n",
    "    if (ae := event.agent_event) and isinstance(ae.step, Respond):\n",
    "        final_tape = ae.partial_tape\n",
    "        break\n",
    "assert final_tape\n",
    "HTML(render_tape_with_prompts(final_tape, PrettyRenderer()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, let's add the search subagent to a financial analyst agent like the ones in earlier examples. We need to give the root agent a way to use the LLM to decide whether to call the search specialist, and if yes, what query to pass to it. We will abuse the tool calling mechanism for this purpose to make the example simpler."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2025-02-24T13:11:39.307928Z",
     "iopub.status.busy": "2025-02-24T13:11:39.307798Z",
     "iopub.status.idle": "2025-02-24T13:12:12.215148Z",
     "shell.execute_reply": "2025-02-24T13:12:12.214502Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style>.basic-renderer-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; white-space: pre-wrap; color: black;}.basic-prompt-box { margin: 4px; padding: 2px; padding-left: 6px; background: lavender; color: black;}.episode-row { display: flex; align-items: end; }.agent-column { width: 50%; }.user-column { width: 25%; }.annotator-column { width: 25%; }.inner-tape-container { display: flex }.inner-tape-indent { width: 10%; }.inner-tape { width: 90%; }.agent_node { text-align: center; position: relative; margin: 12px 0; }.agent_node hr { border: 1px solid #e1e1e1; margin: 0 !important;}.agent_node span { position: absolute; right: 0; top: -0.6em; background: white; color: grey !important; padding: 0 10px; }</style><style>.observation { background-color: #baffc9;; }.error_observation { background-color: #dd0000; }.action { background-color: #cccccc; }.thought { background-color: #ffffdb; }.call { background-color: #ffffff; }.respond { background-color: #ffffff; }.step-header { margin: 2pt 2pt 2pt 0 !important; }.step-text { font-size: 12px; white-space: pre-wrap; word-wrap: break-word;}</style><h3 style='margin: 2px'>Metadata</h3><details> <summary>id: 543f2367-a8c4-493d-8a64-27af3e809a40</summary> <div class='basic-renderer-box'>author: analyst\n",
       "author_tape_id: null\n",
       "error: null\n",
       "id: 543f2367-a8c4-493d-8a64-27af3e809a40\n",
       "n_added_steps: 2\n",
       "parent_id: a9e82ce5-b448-4dde-8cac-dfa30233589a\n",
       "result: {}\n",
       "</div> </details><h3 style='margin: 2px'>Steps</h3><div class='basic-renderer-box observation'><h4 class='step-header'>[0] User</h4><pre class='step-text'>kind: user\n",
       "\n",
       "Tell me about Vulcan in 3 sentences</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 109 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_ticker\",\n",
       "      \"description\": \"Get company stock ticker from its name.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"company_name\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"company_name\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_data\",\n",
       "      \"description\": \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"symbol\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"start_date\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"end_date\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"symbol\",\n",
       "          \"start_date\",\n",
       "          \"end_date\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"call_search_agent\",\n",
       "      \"description\": \"Use this tool to ask a fellow AI agent to search for information on the web.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "system: \n",
       "You will help the user to learn about financials of companies. \n",
       "For general user queries, include some info about stock price changes during the last year, as well as some general information on the company.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "user: Write a natural language plan on how to use tools help the user. Output a list of numbered items, like 1., 2., 3., etc.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 119 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=\"1. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan.\\n2. Use the `functions.call_search_agent` tool to gather general information about Vulcan, including its business operations and industry.\\n3. Use the `functions.get_stock_data` tool to retrieve Vulcan's stock price data for the past year, using the stock ticker obtained in step 1.\\n4. Compile the information from steps 2 and 3 into a concise summary about Vulcan, including stock price changes over the last year and general company information.\", role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box thought'><h4 class='step-header'>[1] Thought: AssistantThought</h4><pre class='step-text'>kind: assistant_thought\n",
       "\n",
       "1. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan.\n",
       "2. Use the `functions.call_search_agent` tool to gather general information about Vulcan, including its business operations and industry.\n",
       "3. Use the `functions.get_stock_data` tool to retrieve Vulcan's stock price data for the past year, using the stock ticker obtained in step 1.\n",
       "4. Compile the information from steps 2 and 3 into a concise summary about Vulcan, including stock price changes over the last year and general company information.</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 213 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_ticker\",\n",
       "      \"description\": \"Get company stock ticker from its name.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"company_name\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"company_name\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_data\",\n",
       "      \"description\": \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"symbol\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"start_date\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"end_date\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"symbol\",\n",
       "          \"start_date\",\n",
       "          \"end_date\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"call_search_agent\",\n",
       "      \"description\": \"Use this tool to ask a fellow AI agent to search for information on the web.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "system: \n",
       "You will help the user to learn about financials of companies. \n",
       "For general user queries, include some info about stock price changes during the last year, as well as some general information on the company.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan.\n",
       "2. Use the `functions.call_search_agent` tool to gather general information about Vulcan, including its business operations and industry.\n",
       "3. Use the `functions.get_stock_data` tool to retrieve Vulcan's stock price data for the past year, using the stock ticker obtained in step 1.\n",
       "4. Compile the information from steps 2 and 3 into a concise summary about Vulcan, including stock price changes over the last year and general company information.\n",
       "--\n",
       "user: Follow the plan you created to earlier. When you are done, respond to the user.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 0 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=None, role='assistant', tool_calls=[ChatCompletionMessageToolCall(function=Function(arguments='{\"company_name\":\"Vulcan\"}', name='get_stock_ticker'), id='call_MnfVd4XRZFgrixiBPT7QAvrn', type='function')], function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box thought'><h4 class='step-header'>[2] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box action'><h4 class='step-header'>[3] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_ticker\n",
       "    arguments: '{\"company_name\":\"Vulcan\"}'\n",
       "  id: call_MnfVd4XRZFgrixiBPT7QAvrn\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[4] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_MnfVd4XRZFgrixiBPT7QAvrn\n",
       "\n",
       "VMC</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 207 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_ticker\",\n",
       "      \"description\": \"Get company stock ticker from its name.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"company_name\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"company_name\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_data\",\n",
       "      \"description\": \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"symbol\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"start_date\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"end_date\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"symbol\",\n",
       "          \"start_date\",\n",
       "          \"end_date\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"call_search_agent\",\n",
       "      \"description\": \"Use this tool to ask a fellow AI agent to search for information on the web.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "system: \n",
       "You will help the user to learn about financials of companies. \n",
       "For general user queries, include some info about stock price changes during the last year, as well as some general information on the company.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan.\n",
       "2. Use the `functions.call_search_agent` tool to gather general information about Vulcan, including its business operations and industry.\n",
       "3. Use the `functions.get_stock_data` tool to retrieve Vulcan's stock price data for the past year, using the stock ticker obtained in step 1.\n",
       "4. Compile the information from steps 2 and 3 into a concise summary about Vulcan, including stock price changes over the last year and general company information.\n",
       "--\n",
       "assistant: [{'function': {'name': 'get_stock_ticker', 'arguments': '{\"company_name\":\"Vulcan\"}'}, 'id': 'call_MnfVd4XRZFgrixiBPT7QAvrn', 'type': 'function'}]\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "user: Follow the plan you created to earlier. When you are done, respond to the user.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 0 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=None, role='assistant', tool_calls=[ChatCompletionMessageToolCall(function=Function(arguments='{\"query\": \"Vulcan company overview\"}', name='call_search_agent'), id='call_EqmIkuSShWQMOddXcGKa8nHM', type='function'), ChatCompletionMessageToolCall(function=Function(arguments='{\"symbol\": \"VMC\", \"start_date\": \"2023-09-17\", \"end_date\": \"2024-09-17\"}', name='get_stock_data'), id='call_k0R8ELYZMVJky1yOVzWlw0WT', type='function')], function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box thought'><h4 class='step-header'>[5] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box call'><h4 class='step-header'>[6] analyst calls search_agent</h4><pre class='step-text'>kind: call\n",
       "agent_name: search_agent\n",
       "\n",
       "Vulcan company overview</pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 17 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"web_search_tool\",\n",
       "      \"description\": \"Search the web for a given query, return a list of search result dictionaries.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"max_results\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"retry_pause\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"attempts\": {\n",
       "            \"default\": 3,\n",
       "            \"type\": \"integer\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_page\",\n",
       "      \"description\": \"Loads a web page and returns a tuple containing:\\n- Content of its first viewport (first screen)\\n- Current page number\\n- Total number of pages\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"url\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"url\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_next_page\",\n",
       "      \"description\": \"Load next page of the document and return the current content of the viewport, current page number and total number of pages.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {},\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "user (analyst): Vulcan company overview</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 0 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=None, role='assistant', tool_calls=[ChatCompletionMessageToolCall(function=Function(arguments='{\"query\":\"Vulcan company overview\"}', name='web_search_tool'), id='call_8A1OQ1gLv3blRmsKEJOP7FxQ', type='function')], function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[7] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: web_search_tool\n",
       "    arguments: '{\"query\":\"Vulcan company overview\"}'\n",
       "  id: call_8A1OQ1gLv3blRmsKEJOP7FxQ\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[8] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: main\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[9] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_8A1OQ1gLv3blRmsKEJOP7FxQ\n",
       "\n",
       "<details><summary>1390 characters ...</summary>[{'title': 'About Vulcan', 'url': 'https://www.vulcanmaterials.com/about-vulcan', 'content': \"Vulcan Materials Company is the nation's largest producer of construction aggregates—primarily crushed stone, sand and gravel—and a major producer of aggregates ...\"}, {'title': 'Vulcan Materials Company', 'url': 'https://www.vulcanmaterials.com/', 'content': \"Vulcan Materials Company is the nation's largest producer of construction aggregates—primarily crushed stone, sand and gravel ...\"}, {'title': 'Vulcan Materials Co - Company Profile - GlobalData', 'url': 'https://www.globaldata.com/company-profile/vulcan-materials-company/', 'content': 'Vulcan Materials Co (Vulcan) is a manufacturer of construction materials. It sells crushed stone, sand, gravel and aggregates-based building materials.'}, {'title': 'Vulcan Materials | VMC Stock Price, Company Overview & News', 'url': 'https://www.forbes.com/companies/vulcan-materials/', 'content': 'Vulcan Materials Co. engages in the provision of basic materials and supply for infrastructure and construction industry. It operates through the following.'}, {'title': 'Vulcan Materials Company - Wikipedia', 'url': 'https://en.wikipedia.org/wiki/Vulcan_Materials_Company', 'content': 'Vulcan is the largest producer of construction materials, primarily gravel, crushed stone, and sand, and employs approximately 12,000 people at over 400 ...'}]</details></pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 337 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"web_search_tool\",\n",
       "      \"description\": \"Search the web for a given query, return a list of search result dictionaries.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"max_results\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"retry_pause\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"attempts\": {\n",
       "            \"default\": 3,\n",
       "            \"type\": \"integer\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_page\",\n",
       "      \"description\": \"Loads a web page and returns a tuple containing:\\n- Content of its first viewport (first screen)\\n- Current page number\\n- Total number of pages\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"url\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"url\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_next_page\",\n",
       "      \"description\": \"Load next page of the document and return the current content of the viewport, current page number and total number of pages.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {},\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "user (analyst): Vulcan company overview\n",
       "--\n",
       "assistant: [{'function': {'name': 'web_search_tool', 'arguments': '{\"query\":\"Vulcan company overview\"}'}, 'id': 'call_8A1OQ1gLv3blRmsKEJOP7FxQ', 'type': 'function'}]\n",
       "--\n",
       "tool: [{'title': 'About Vulcan', 'url': 'https://www.vulcanmaterials.com/about-vulcan', 'content': \"Vulcan Materials Company is the nation's largest producer of construction aggregates—primarily crushed stone, sand and gravel—and a major producer of aggregates ...\"}, {'title': 'Vulcan Materials Company', 'url': 'https://www.vulcanmaterials.com/', 'content': \"Vulcan Materials Company is the nation's largest producer of construction aggregates—primarily crushed stone, sand and gravel ...\"}, {'title': 'Vulcan Materials Co - Company Profile - GlobalData', 'url': 'https://www.globaldata.com/company-profile/vulcan-materials-company/', 'content': 'Vulcan Materials Co (Vulcan) is a manufacturer of construction materials. It sells crushed stone, sand, gravel and aggregates-based building materials.'}, {'title': 'Vulcan Materials | VMC Stock Price, Company Overview & News', 'url': 'https://www.forbes.com/companies/vulcan-materials/', 'content': 'Vulcan Materials Co. engages in the provision of basic materials and supply for infrastructure and construction industry. It operates through the following.'}, {'title': 'Vulcan Materials Company - Wikipedia', 'url': 'https://en.wikipedia.org/wiki/Vulcan_Materials_Company', 'content': 'Vulcan is the largest producer of construction materials, primarily gravel, crushed stone, and sand, and employs approximately 12,000 people at over 400 ...'}]</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 0 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=None, role='assistant', tool_calls=[ChatCompletionMessageToolCall(function=Function(arguments='{\"url\": \"https://www.vulcanmaterials.com/about-vulcan\"}', name='get_page'), id='call_sjatF0zDhuRCSFsECuQsfN6R', type='function'), ChatCompletionMessageToolCall(function=Function(arguments='{\"url\": \"https://en.wikipedia.org/wiki/Vulcan_Materials_Company\"}', name='get_page'), id='call_UVgKW4ijEx7Hm46QFUxeP6E2', type='function')], function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box action'><h4 class='step-header'>[10] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_page\n",
       "    arguments: '{\"url\": \"https://www.vulcanmaterials.com/about-vulcan\"}'\n",
       "  id: call_sjatF0zDhuRCSFsECuQsfN6R\n",
       "  type: function\n",
       "- function:\n",
       "    name: get_page\n",
       "    arguments: '{\"url\": \"https://en.wikipedia.org/wiki/Vulcan_Materials_Company\"}'\n",
       "  id: call_UVgKW4ijEx7Hm46QFUxeP6E2\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box thought'><h4 class='step-header'>[11] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: main\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[12] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_sjatF0zDhuRCSFsECuQsfN6R\n",
       "\n",
       "<details><summary>11802 characters ...</summary>('Title: \\n\\tAbout Vulcan\\n\\n=======================\\n[![Vulcan Materials Company](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/vulcan-materials-logo.png?sfvrsn=c2926ad1_1 \"Vulcan Materials Company\")](https://www.vulcanmaterials.com/)\\n\\n* [Careers](/careers)\\n* [News Releases](http://ir.vulcanmaterials.com/investor-relations/news-releases/default.aspx)\\n* [Employees](https://vul.onelogin.com)\\n* [Contact](/contact)\\n\\n| [Vulcan\\'s Statement on Inclusion \\\\& Diversity](https://csr.vulcanmaterials.com/2020/07/diversity-statement/) [Website Accessibility Statement](/accessibility-statement) |\\n| --- |\\n\\n[![Facebook](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/f_logo_rgb-hex-blue_512.png?sfvrsn=effc00bb_1)](https://www.facebook.com/VulcanMaterialsCompany/)\\xa0\\xa0[![Instagram](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/instagram_appicon_aug2017.png?sfvrsn=b5c79c26_1)](https://www.instagram.com/vulcanmaterialscompany/)\\xa0\\xa0[![YouTube](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/youtube-35px.png?sfvrsn=4dffab8d_1)](https://www.youtube.com/user/VulcanCorporate/videos)\\xa0\\xa0[![LinkedIn](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/linkedin-35px.png?sfvrsn=c4e982b3_1)](https://www.linkedin.com/company/vulcan-materials-company/)\\n\\n* [About Vulcan](https://www.vulcanmaterials.com/about-vulcan)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/navigation-images/about-vulcan.png?sfvrsn=bb5ab483_1)\\n\\n\\t+ [History](https://www.vulcanmaterials.com/about-vulcan/history)\\n\\t+ [Mission and Values](https://www.vulcanmaterials.com/about-vulcan/mission-and-values)\\n\\t+ [Executive Management](https://www.vulcanmaterials.com/about-vulcan/executive-management)\\n\\t+ [Corporate Office](https://www.vulcanmaterials.com/about-vulcan/corporate-office)\\n\\t+ [Industry Links](https://www.vulcanmaterials.com/about-vulcan/industry-links)\\n\\t+ [Careers](https://www.vulcanmaterials.com/about-vulcan/careers)\\n* [Construction Materials](https://www.vulcanmaterials.com/construction-materials)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/navigation-images/construction-materials.png?sfvrsn=e2b2e8f0_1)\\n\\n\\t+ [MyVulcan](https://www.vulcanmaterials.com/construction-materials/myvulcan)\\n\\t+ [Product Sales](https://www.vulcanmaterials.com/construction-materials/product-sales)\\n\\t+ [Products and Services](https://www.vulcanmaterials.com/construction-materials/products-and-services)\\n\\t+ [Suppliers](https://www.vulcanmaterials.com/construction-materials/suppliers)\\n\\t+ [Facilities Map](https://www.vulcanmaterials.com/construction-materials/facilities-map)\\n\\t+ [Operating Groups](https://www.vulcanmaterials.com/construction-materials/operating-groups)\\n\\t+ [Product Calculators](https://www.vulcanmaterials.com/construction-materials/product-calculators)\\n\\t+ [Credit Applications](https://www.vulcanmaterials.com/construction-materials/credit-applications)\\n\\t+ [Safety Data Sheets](https://www.vulcanmaterials.com/construction-materials/safety-data-sheets)\\n\\t+ [Product Declaration](https://www.vulcanmaterials.com/construction-materials/product-declaration)\\n\\t+ [Facilities](https://www.vulcanmaterials.com/construction-materials/facilities)\\n* [Investor Relations](https://www.vulcanmaterials.com/investor-relations)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/navigation-images/thmb-overview.jpg?sfvrsn=bfa2dc74_1)\\n\\n\\t+ [Overview](https://www.vulcanmaterials.com/investor-relations/overview)\\n\\t+ [Why Invest](https://www.vulcanmaterials.com/investor-relations/why-invest)\\n\\t+ [News](https://www.vulcanmaterials.com/investor-relations/news)\\n\\t+ [Events \\\\& Presentations](https://www.vulcanmaterials.com/investor-relations/events-presentations)\\n\\t+ [Stock Info](https://www.vulcanmaterials.com/investor-relations/stock-info)\\n\\t+ [Financials](https://www.vulcanmaterials.com/investor-relations/financials)\\n\\t+ [Governance](https://www.vulcanmaterials.com/investor-relations/governance)\\n\\t+ [Link to Earnings Call](https://www.vulcanmaterials.com/investor-relations/link-to-earnings-call)\\n\\t+ [Link to 2022 Investor Day](https://www.vulcanmaterials.com/investor-relations/link-to-2022-investor-day)\\n* [Social Responsibility](https://www.vulcanmaterials.com/social-responsibility)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/navigation-images/social-responsibility.jpg?sfvrsn=189cf41b_1)\\n\\n\\t+ [2023 Sustainability Report](https://www.vulcanmaterials.com/social-responsibility/2023-sustainability-report)\\n\\t+ [Safety \\\\& Health](https://www.vulcanmaterials.com/social-responsibility/safety-health)\\n\\t+ [Environmental Stewardship](https://www.vulcanmaterials.com/social-responsibility/environmental-stewardship)\\n\\t+ [People](https://www.vulcanmaterials.com/social-responsibility/people)\\n\\t+ [Community](https://www.vulcanmaterials.com/social-responsibility/community)\\n\\t+ [Governance](https://www.vulcanmaterials.com/social-responsibility/governance)\\n\\t+ [Teacher Center](https://www.vulcanmaterials.com/social-responsibility/teacher-center)\\n\\t+ [Vulcan Foundation](https://www.vulcanmaterials.com/social-responsibility/vulcan-foundation)\\n\\nAbout Vulcan\\n============\\n\\n* [About Vulcan](https://www.vulcanmaterials.com/about-vulcan)\\n\\n\\t+ [History](https://www.vulcanmaterials.com/about-vulcan/history)\\n\\t+ [Mission and Values](https://www.vulcanmaterials.com/about-vulcan/mission-and-values)\\n\\t+ [Executive Management](https://www.vulcanmaterials.com/about-vulcan/executive-management)\\n\\t+ [Corporate Office](https://www.vulcanmaterials.com/about-vulcan/corporate-office)\\n\\t+ [Industry Links](https://www.vulcanmaterials.com/about-vulcan/industry-links)\\n\\t+ [Careers](https://www.vulcanmaterials.com/about-vulcan/careers)\\n\\n* [Home](https://www.vulcanmaterials.com/)\\n\\\\>\\n* About Vulcan\\n\\nOur business is built on essentials. What we mine and sell—aggregates—are indispensable to infrastructure and growth. How we run our company also comes down to essentials. We run our business with great discipline and integrity. We maintain a great place to work and grow. We take care of the land through the entire cycle of productive use from mining to restoration and re\\\\-use. We serve our customers and communities well. Always with the long view in mind. This is how we sustain success.\\n\\nWe are Vulcan. And we deliver value.\\n\\n![About Vulcan](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/about-vulcan/aboutlanding_body.png?sfvrsn=c96e2a2f_1)\\n\\nVulcan Materials Company is the nation’s largest producer of construction aggregates—primarily crushed stone, sand and gravel—and a major producer of aggregates\\\\-based construction materials including asphalt and ready\\\\-mixed concrete. Our coast\\\\-to\\\\-coast footprint and strategic distribution network align with and serve the nation’s growth centers. We are headquartered in Birmingham, Alabama.\\n\\nVulcan has over 397\\\\*\\xa0active aggregates facilities, and 66\\\\*\\xa0asphalt facilities and 63\\\\*\\xa0concrete facilities, which also consume aggregates.\\xa0Most\\xa0are located in the U.S. except for our quarries and marine terminals on Vancouver Island in British Columbia, Canada,\\xa0Honduras\\xa0and Mexico’s Yucatán Peninsula.\\n\\nWhat we produce is used in nearly all forms of construction. In particular, large quantities of aggregates are used to build and repair valuable infrastructure such as roads, bridges, waterworks and ports, and to construct buildings both residential and nonresidential, such as manufacturing facilities, office buildings, schools, hospitals and churches.\\n\\n\\xa0\\\\* Data from 2023 10k and Annual Report\\n\\n[Meet Vulcan\\'s Leadership](https://ir.vulcanmaterials.com/governance/executive-management/default.aspx)\\n\\n[Vulcan Corporate Office](/about-vulcan/corporate-office)\\n\\n[Vulcan History](/about-vulcan/history)\\n\\n[Industry Links](/about-vulcan/industry-links)\\n\\n### [Contact](/contact)\\n\\n**Vulcan Materials Company**\\nCorporate Office\\n1200 Urban Center Drive\\nBirmingham, AL 35242\\nTel: (205\\\\) 298\\\\-3000\\n\\nPlease follow the hyperlinks to review the [Business Conduct Policy](https://s3.us-east-2.amazonaws.com/files.vulcanmaterials.com/vmc/Business-Conduct-Policy.pdf) or to report [concerns](https://secure.ethicspoint.com/domain/media/en/gui/16082/index.html).\\n\\nSiga los hipervínculos para revisar la\\xa0[Política de Conducta Empresarial](https://s3.us-east-2.amazonaws.com/files.vulcanmaterials.com/vmc/VMC-Business_Conduct_Policy_Spanish-2021.pdf)\\xa0o para reportar algún\\xa0[problema](https://secure.ethicspoint.com/domain/media/es/gui/16082/index.html).\\n\\nContact us for\\xa0[Product Sales](/construction-materials/product-sales).\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/footer_aboutvulcan.png?sfvrsn=a4a41105_1)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/footer_constructionmaterials.png?sfvrsn=c59fead5_1)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/footer_investorrelations.png?sfvrsn=3d684e71_1)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/social-responsibility-footer.jpg?sfvrsn=c6651106_1)\\n\\nAbout Vulcan\\n\\n* [History](/about-vulcan/history)\\n* [Mission and Values](/about-vulcan/mission-and-values)\\n* [Executive Management](/about-vulcan/executive-management)\\n* [Corporate Office](/about-vulcan/corporate-office)\\n* [Industry Links](/about-vulcan/industry-links)\\n* [Careers](/about-vulcan/careers)\\n\\nConstruction Materials\\n\\n* [MyVulcan](/construction-materials/myvulcan)\\n* [Product Sales](/construction-materials/product-sales)\\n* [Products and Services](/construction-materials/products-and-services)\\n* [Suppliers](/construction-materials/suppliers)\\n* [Facilities Map](/construction-materials/facilities-map)\\n* [Operating Groups](/construction-materials/operating-groups)\\n* [Product Calculators](/construction-materials/product-calculators)\\n* [Credit Applications](/construction-materials/credit-applications)\\n* [Safety Data Sheets](/construction-materials/safety-data-sheets)\\n* [Product Declaration](/construction-materials/product-declaration)\\n* [Facilities](/construction-materials/facilities)\\n\\nInvestor Relations\\n\\n* [Overview](/investor-relations/overview)\\n* [Why Invest](/investor-relations/why-invest)\\n* [News](/investor-relations/news)\\n* [Events \\\\& Presentations](/investor-relations/events-presentations)\\n* [Stock Info](/investor-relations/stock-info)\\n* [Financials](/investor-relations/financials)\\n* [Governance](/investor-relations/governance)\\n* [Link to Earnings Call](/investor-relations/link-to-earnings-call)\\n* [Link to 2022 Investor Day](/investor-relations/link-to-2022-investor-day)\\n\\nSocial Responsibility\\n\\n* [2023 Sustainability Report](/social-responsibility/2023-sustainability-report)\\n* [Safety \\\\& Health](/social-responsibility/safety-health)\\n* [Environmental Stewardship](/social-responsibility/environmental-stewardship)\\n* [People](/social-responsibility/people)\\n* [Community](/social-responsibility/community)\\n* [Governance](/social-responsibility/governance)\\n* [Teacher Center](/social-responsibility/teacher-center)\\n* [Vulcan Foundation](/social-responsibility/vulcan-foundation)\\n\\n* [Site Map](/site-map)\\n* [Terms of Use](/terms-of-use)\\n* [Privacy Policy](/privacy-policy)\\n* [California Residents\\' Privacy Notice](/california-privacy-notice)\\n* [Contact Us About Our Site](/contact)\\n© 2024 Vulcan Materials Company', 1, 0)</details></pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[13] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_UVgKW4ijEx7Hm46QFUxeP6E2\n",
       "\n",
       "<details><summary>10857 characters ...</summary>('Title: Vulcan Materials Company - Wikipedia\\n=======================\\n# Vulcan Materials Company\\n\\nAmerican construction materials company\\n\\nVulcan Materials Company\\n|  | |\\n| --- | --- |\\n| Company type | Public company |\\n| [Traded as](/wiki/Ticker_symbol \"Ticker symbol\") | [NYSE](/wiki/New_York_Stock_Exchange \"New York Stock Exchange\"):\\xa0[VMC](https://www.nyse.com/quote/XNYS:VMC)[S\\\\&P 500 Component](/wiki/S%26P_500 \"S&P 500\") |\\n| Industry | [Construction materials](/wiki/Construction_materials \"Construction materials\") |\\n| Founded | 1909 as **Birmingham Slag Company** |\\n| Headquarters | [Birmingham](/wiki/Birmingham,_Alabama \"Birmingham, Alabama\"), [Alabama](/wiki/Alabama \"Alabama\") [United States](/wiki/United_States \"United States\") |\\n| Key people | [J. Thomas Hill](/wiki/J._Thomas_Hill \"J. Thomas Hill\"), [Chief Executive Officer](/wiki/Chief_Executive_Officer \"Chief Executive Officer\")Suzanne H. Wood, [Chief Financial Officer](/wiki/Chief_Financial_Officer \"Chief Financial Officer\")Denson N. Franklin III, [General Counsel](/wiki/General_Counsel \"General Counsel\") |\\n| Products | Crushed stone, sand and gravel |\\n| Revenue | Increase [$](/wiki/United_States_Dollar \"United States Dollar\")7\\\\.315 billion (2022\\\\)[\\\\[1]](#cite_note-Vulcan_2022-1) |\\n| Number of employees | 12,000[\\\\[2]](#cite_note-2)\\xa0(2023\\\\) |\\n| Website | [VulcanMaterials.com](https://www.vulcanmaterials.com) |\\n\\n**Vulcan Materials Company** ([NYSE](/wiki/New_York_Stock_Exchange \"New York Stock Exchange\"):\\xa0[VMC](https://www.nyse.com/quote/XNYS:VMC)) is an American company based in [Birmingham](/wiki/Birmingham,_Alabama \"Birmingham, Alabama\"), [Alabama](/wiki/Alabama \"Alabama\"). It is principally engaged in the production, distribution and sale of construction materials. Vulcan is the largest producer of construction materials, primarily gravel, crushed stone, and sand, and employs approximately 12,000 people at over 400 facilities. Vulcan serves 22 states, the [District of Columbia](/wiki/District_of_Columbia \"District of Columbia\"), [Mexico](/wiki/Mexico \"Mexico\"), [Canada](/wiki/Canada \"Canada\"), [Bahamas](/wiki/Bahamas \"Bahamas\") and the [U.S. Virgin Islands](/wiki/U.S._Virgin_Islands \"U.S. Virgin Islands\"). Vulcan\\'s innovative [Crescent Market project](/w/index.php?title=Crescent_Market_project&action=edit&redlink=1 \"Crescent Market project (page does not exist)\") led to construction of a large quarry and deep water seaport on the [Yucatán Peninsula](/wiki/Yucat%C3%A1n_Peninsula \"Yucatán Peninsula\") of Mexico, just south of [Cancun](/wiki/Cancun \"Cancun\"). This quarry supplies Tampa, New Orleans, Houston, and Brownsville, Texas, as well as other Gulf coast seaports, with crushed limestone via large 62,000\\\\-ton self\\\\-discharging ships.\\n\\nHistory\\n-------\\n\\n\\\\[[edit](/w/index.php?title=Vulcan_Materials_Company&action=edit&section=1 \"Edit section: History\")]\\nIn the late 1980s Vulcan acquired Reed Crushed Stone, near [Kentucky](/wiki/Gravel_Switch,_Livingston_County,_Kentucky \"Gravel Switch, Livingston County, Kentucky\").[\\\\[3]](#cite_note-3)[\\\\[4]](#cite_note-4)\\n\\nOn November 16, 1998, Vulcan announced it was purchasing CalMat Company, a producer of asphalt and ready\\\\-mixed concrete based in [Los Angeles](/wiki/Los_Angeles \"Los Angeles\"), for $760 million in cash.[\\\\[5]](#cite_note-5)\\n\\nOn June 7, 2005, Vulcan completed the sale of its chemicals business, known as Vulcan Chemicals, to [Occidental Chemical Corporation](/wiki/Occidental_Chemical_Corporation \"Occidental Chemical Corporation\"). The sale of assets included Vulcan\\'s [chloralkali](/wiki/Chloralkali_process \"Chloralkali process\") plants in Wichita, Kansas; Geismar, Louisiana; and Port Edwards, Wisconsin, as well as Vulcan Chemicals\\' joint venture located in Geismar, Louisiana. In 2005, Vulcan acquired 11 aggregates operations and five asphalt plants in Arizona, Georgia, Indiana and Tennessee.\\n\\nOn February 19, 2007, Vulcan announced that it would buy stone and cement producer Florida Rock Industries for $4\\\\.7 billion.[\\\\[6]](#cite_note-6)[\\\\[7]](#cite_note-7) Vulcan completed the acquisition of Florida Rock on November 16, 2007\\\\.[\\\\[8]](#cite_note-8)\\n\\nIn March 2007, Vulcan announced that it had been named to [Fortune Magazine](/wiki/Fortune_Magazine \"Fortune Magazine\")\\'s list of *Most Admired Companies* for the sixth time. The company was ranked first in its industry sector, \"Building Materials, Glass.\" Overall, Vulcan ranked among the top 10 companies in the [Fortune 1000](/wiki/Fortune_1000 \"Fortune 1000\") for both long\\\\-term investment and [social responsibility](/wiki/Social_responsibility \"Social responsibility\").\\n\\nIn January 2014, Vulcan Materials announced they would be selling their Florida Rock Industry (cement and ready\\\\-mixed concrete) plants and equipment to [Cementos Argos](/wiki/Argos_S.A. \"Argos S.A.\"), of Colombia. In March 2014, Vulcan was able to fully divest these properties to Cementos Argos.[\\\\[9]](#cite_note-FloridaRockSale-9)\\n\\nIn January 2018, Vulcan Materials was found to be partially responsible for the [Bayou Corne sinkhole](/wiki/Bayou_Corne_sinkhole \"Bayou Corne sinkhole\"), along with [Occidental Petroleum](/wiki/Occidental_Petroleum \"Occidental Petroleum\") and Texas Brine Company.[\\\\[10]](#cite_note-10)\\n\\nIn June 2021, Vulcan Materials announced they would be acquiring [US Concrete (USCR)](/wiki/U.S._Concrete \"U.S. Concrete\") for $74 per share, a 30% premium on the date of the announcement. U.S. Concrete was formed in 1999, with its headquarters in Euless, TX. The deal is valued at $1\\\\.3 billion. The acquisition will allow for integrated expansion of Vulcan and make the subsidiary debt\\\\-free with adequate working capital. The deal has been approved by both boards and will close in the second half of 2021\\\\.[\\\\[11]](#cite_note-USConcrete_Acquisition-11)\\n\\nIn early 2023 after being accused by president [Andrés Manuel López Obrador](/wiki/L%C3%B3pez_Obrador \"López Obrador\") of exporting materials from Mexico without a permit, the Mexican military seized a marine terminal belonging to Vulcan leading to a diplomatic spat between the US and Mexico.[\\\\[12]](#cite_note-12)\\n\\nReferences\\n----------\\n\\n\\\\[[edit](/w/index.php?title=Vulcan_Materials_Company&action=edit&section=2 \"Edit section: References\")]\\n\\n1. **[^](#cite_ref-Vulcan_2022_1-0)** [\"VULCAN REPORTS FOURTH QUARTER AND FULL YEAR 2022 RESULTS\"](https://ir.vulcanmaterials.com/news/news-details/2023/VULCAN-REPORTS-FOURTH-QUARTER-AND-FULL-YEAR-2022-RESULTS/default.aspx). Vulcan Materials Company. Retrieved Apr 2, 2023.\\n2. **[^](#cite_ref-2)** [\"Vulcan Materials\"](https://www.sec.gov/ix?doc=/Archives/edgar/data/1396009/000139600923000007/vmc-20221231x10k.htm#PartII_Item8). *SEC.gov*. Retrieved 2023\\\\-04\\\\-02.\\n3. **[^](#cite_ref-3)** [*Minerals Yearbook*](https://books.google.com/books?id=WkoZSm8sXQIC&q=Reed+Terminal+aggregates&pg=PA211). 1989\\\\. [ISBN](/wiki/ISBN_(identifier) \"ISBN (identifier)\")\\xa0[9780160359316](/wiki/Special:BookSources/9780160359316 \"Special:BookSources/9780160359316\").\\n4. **[^](#cite_ref-4)** [*Minerals Yearbook*](https://books.google.com/books?id=EDW2hvyndDUC&q=vulcan+mineral+grand+river+operations&pg=PA191). 2009\\\\. [ISBN](/wiki/ISBN_(identifier) \"ISBN (identifier)\")\\xa0[9781411323292](/wiki/Special:BookSources/9781411323292 \"Special:BookSources/9781411323292\").\\n5. **[^](#cite_ref-5)** Vrana, Debora (November 16, 1998\\\\). [\"Concrete Firm CalMat to Be Acquired for $760 Million\"](https://www.latimes.com/archives/la-xpm-1998-nov-16-fi-43297-story.html). *The Los Angeles Times*. Retrieved 2024\\\\-08\\\\-10.\\n6. **[^](#cite_ref-6)** Dolmetsch, Chris \\\\& Lococo, Edmond (19 Feb 2007\\\\). [\"Vulcan Materials to Buy Florida Rock for $4\\\\.6 Billion (Update4\\\\)\"](https://www.bloomberg.com/apps/news?pid=newsarchive&sid=amqbTdNtr8Pw). *Bloomberg*. Retrieved 12 May 2015.\\n7. **[^](#cite_ref-7)** [\"Vulcan Materials to buy Florida Rock for $4\\\\.6 billion\"](https://www.reuters.com/article/us-floridarock-vulcan-idUSN1944560620070220). *[Reuters](/wiki/Reuters \"Reuters\")*. 19 February 2007. Retrieved 27 May 2021.\\n8. **[^](#cite_ref-8)** Zubko, Nick (19 March 2008\\\\). [\"Florida Rock Industries Inc.: Under New Management\"](https://www.industryweek.com/leadership/companies-executives/article/21942309/florida-rock-industries-inc-under-new-management). *Industry Week*. Retrieved 27 May 2021.\\n9. **[^](#cite_ref-FloridaRockSale_9-0)** Basch, Mark (27 January 2014\\\\). [\"Vulcan selling former Florida Rock cement facilities\"](https://www.jaxdailyrecord.com/showstory.php?Story_id=541695). *Jacksonville Daily Record*. Retrieved 12 May 2015.\\n10. **[^](#cite_ref-10)** [Judge: Fault for Bayou Corne sinkhole lies with Texas Brine, OxyChem, Vulcan; companies had decades of warnings](https://www.theadvocate.com/baton_rouge/news/article_7856ff5e-f4ae-11e7-b86c-4f261682612b.html) [The Advocate](/wiki/The_Advocate_(Louisiana) \"The Advocate (Louisiana)\")\\n11. **[^](#cite_ref-USConcrete_Acquisition_11-0)** Lenihan, Rob (7 June 2021\\\\). [\"U.S. Concrete Skyrockets on Deal to Be Bought for $1\\\\.29B by Vulcan\"](https://www.thestreet.com/investing/us-concrete-uscr-vulcan-materials-vmc-acquisition). *The Street*. Retrieved 7 Jun 2021.\\n12. **[^](#cite_ref-12)** Haldevang, Max (19 Mar 2023\\\\). [\"Vulcan Facility Seizure Adds to Tension Between US and Mexico\"](https://www.bloomberg.com/news/articles/2023-03-19/vulcan-facility-seizure-adds-to-tension-between-us-and-mexico?leadSource=uverify%20wall).\\n\\nExternal links\\n--------------\\n\\n\\\\[[edit](/w/index.php?title=Vulcan_Materials_Company&action=edit&section=3 \"Edit section: External links\")]\\n* [Vulcan Materials website](http://www.vulcanmaterials.com)\\n* [*U.S. v. Vulcan Materials Co. and Florida Rock Industries Inc.*](https://www.justice.gov/atr/case/us-v-vulcan-materials-co-and-florida-rock-industries-inc), a US Supreme court case\\n* Business data for Vulcan Materials Company:\\n\\t+ [Google](https://www.google.com/finance/quote/VMC)\\n\\t+ [SEC filings](https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=VMC)\\n\\t+ [Yahoo!](https://finance.yahoo.com/quote/VMC)\\n\\n| [Authority control databases](/wiki/Help:Authority_control \"Help:Authority control\") [Edit this at Wikidata](https://www.wikidata.org/wiki/Q1685287#identifiers \"Edit this at Wikidata\") | |\\n| --- | --- |\\n| International | * [ISNI](https://isni.org/isni/0000000449084962) * [VIAF](https://viaf.org/viaf/131972586) |\\n| National | * [United States](https://id.loc.gov/authorities/n84041729) |\\n\\n![](https://login.wikimedia.org/wiki/Special:CentralAutoLogin/start?useformat=desktop&type=1x1&usesul3=0)\\nRetrieved from \"[https://en.wikipedia.org/w/index.php?title\\\\=Vulcan\\\\_Materials\\\\_Company\\\\&oldid\\\\=1248703031](https://en.wikipedia.org/w/index.php?title=Vulcan_Materials_Company&oldid=1248703031)\"', 1, 0)</details></pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 6776 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"web_search_tool\",\n",
       "      \"description\": \"Search the web for a given query, return a list of search result dictionaries.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"max_results\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"retry_pause\": {\n",
       "            \"default\": 5,\n",
       "            \"type\": \"integer\"\n",
       "          },\n",
       "          \"attempts\": {\n",
       "            \"default\": 3,\n",
       "            \"type\": \"integer\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_page\",\n",
       "      \"description\": \"Loads a web page and returns a tuple containing:\\n- Content of its first viewport (first screen)\\n- Current page number\\n- Total number of pages\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"url\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"url\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_next_page\",\n",
       "      \"description\": \"Load next page of the document and return the current content of the viewport, current page number and total number of pages.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {},\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "user (analyst): Vulcan company overview\n",
       "--\n",
       "assistant: [{'function': {'name': 'web_search_tool', 'arguments': '{\"query\":\"Vulcan company overview\"}'}, 'id': 'call_8A1OQ1gLv3blRmsKEJOP7FxQ', 'type': 'function'}]\n",
       "--\n",
       "tool: [{'title': 'About Vulcan', 'url': 'https://www.vulcanmaterials.com/about-vulcan', 'content': \"Vulcan Materials Company is the nation's largest producer of construction aggregates—primarily crushed stone, sand and gravel—and a major producer of aggregates ...\"}, {'title': 'Vulcan Materials Company', 'url': 'https://www.vulcanmaterials.com/', 'content': \"Vulcan Materials Company is the nation's largest producer of construction aggregates—primarily crushed stone, sand and gravel ...\"}, {'title': 'Vulcan Materials Co - Company Profile - GlobalData', 'url': 'https://www.globaldata.com/company-profile/vulcan-materials-company/', 'content': 'Vulcan Materials Co (Vulcan) is a manufacturer of construction materials. It sells crushed stone, sand, gravel and aggregates-based building materials.'}, {'title': 'Vulcan Materials | VMC Stock Price, Company Overview & News', 'url': 'https://www.forbes.com/companies/vulcan-materials/', 'content': 'Vulcan Materials Co. engages in the provision of basic materials and supply for infrastructure and construction industry. It operates through the following.'}, {'title': 'Vulcan Materials Company - Wikipedia', 'url': 'https://en.wikipedia.org/wiki/Vulcan_Materials_Company', 'content': 'Vulcan is the largest producer of construction materials, primarily gravel, crushed stone, and sand, and employs approximately 12,000 people at over 400 ...'}]\n",
       "--\n",
       "assistant: [{'function': {'name': 'get_page', 'arguments': '{\"url\": \"https://www.vulcanmaterials.com/about-vulcan\"}'}, 'id': 'call_sjatF0zDhuRCSFsECuQsfN6R', 'type': 'function'}, {'function': {'name': 'get_page', 'arguments': '{\"url\": \"https://en.wikipedia.org/wiki/Vulcan_Materials_Company\"}'}, 'id': 'call_UVgKW4ijEx7Hm46QFUxeP6E2', 'type': 'function'}]\n",
       "--\n",
       "tool: ('Title: \\n\\tAbout Vulcan\\n\\n=======================\\n[![Vulcan Materials Company](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/vulcan-materials-logo.png?sfvrsn=c2926ad1_1 \"Vulcan Materials Company\")](https://www.vulcanmaterials.com/)\\n\\n* [Careers](/careers)\\n* [News Releases](http://ir.vulcanmaterials.com/investor-relations/news-releases/default.aspx)\\n* [Employees](https://vul.onelogin.com)\\n* [Contact](/contact)\\n\\n| [Vulcan\\'s Statement on Inclusion \\\\& Diversity](https://csr.vulcanmaterials.com/2020/07/diversity-statement/) [Website Accessibility Statement](/accessibility-statement) |\\n| --- |\\n\\n[![Facebook](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/f_logo_rgb-hex-blue_512.png?sfvrsn=effc00bb_1)](https://www.facebook.com/VulcanMaterialsCompany/)\\xa0\\xa0[![Instagram](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/instagram_appicon_aug2017.png?sfvrsn=b5c79c26_1)](https://www.instagram.com/vulcanmaterialscompany/)\\xa0\\xa0[![YouTube](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/youtube-35px.png?sfvrsn=4dffab8d_1)](https://www.youtube.com/user/VulcanCorporate/videos)\\xa0\\xa0[![LinkedIn](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/linkedin-35px.png?sfvrsn=c4e982b3_1)](https://www.linkedin.com/company/vulcan-materials-company/)\\n\\n* [About Vulcan](https://www.vulcanmaterials.com/about-vulcan)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/navigation-images/about-vulcan.png?sfvrsn=bb5ab483_1)\\n\\n\\t+ [History](https://www.vulcanmaterials.com/about-vulcan/history)\\n\\t+ [Mission and Values](https://www.vulcanmaterials.com/about-vulcan/mission-and-values)\\n\\t+ [Executive Management](https://www.vulcanmaterials.com/about-vulcan/executive-management)\\n\\t+ [Corporate Office](https://www.vulcanmaterials.com/about-vulcan/corporate-office)\\n\\t+ [Industry Links](https://www.vulcanmaterials.com/about-vulcan/industry-links)\\n\\t+ [Careers](https://www.vulcanmaterials.com/about-vulcan/careers)\\n* [Construction Materials](https://www.vulcanmaterials.com/construction-materials)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/navigation-images/construction-materials.png?sfvrsn=e2b2e8f0_1)\\n\\n\\t+ [MyVulcan](https://www.vulcanmaterials.com/construction-materials/myvulcan)\\n\\t+ [Product Sales](https://www.vulcanmaterials.com/construction-materials/product-sales)\\n\\t+ [Products and Services](https://www.vulcanmaterials.com/construction-materials/products-and-services)\\n\\t+ [Suppliers](https://www.vulcanmaterials.com/construction-materials/suppliers)\\n\\t+ [Facilities Map](https://www.vulcanmaterials.com/construction-materials/facilities-map)\\n\\t+ [Operating Groups](https://www.vulcanmaterials.com/construction-materials/operating-groups)\\n\\t+ [Product Calculators](https://www.vulcanmaterials.com/construction-materials/product-calculators)\\n\\t+ [Credit Applications](https://www.vulcanmaterials.com/construction-materials/credit-applications)\\n\\t+ [Safety Data Sheets](https://www.vulcanmaterials.com/construction-materials/safety-data-sheets)\\n\\t+ [Product Declaration](https://www.vulcanmaterials.com/construction-materials/product-declaration)\\n\\t+ [Facilities](https://www.vulcanmaterials.com/construction-materials/facilities)\\n* [Investor Relations](https://www.vulcanmaterials.com/investor-relations)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/navigation-images/thmb-overview.jpg?sfvrsn=bfa2dc74_1)\\n\\n\\t+ [Overview](https://www.vulcanmaterials.com/investor-relations/overview)\\n\\t+ [Why Invest](https://www.vulcanmaterials.com/investor-relations/why-invest)\\n\\t+ [News](https://www.vulcanmaterials.com/investor-relations/news)\\n\\t+ [Events \\\\& Presentations](https://www.vulcanmaterials.com/investor-relations/events-presentations)\\n\\t+ [Stock Info](https://www.vulcanmaterials.com/investor-relations/stock-info)\\n\\t+ [Financials](https://www.vulcanmaterials.com/investor-relations/financials)\\n\\t+ [Governance](https://www.vulcanmaterials.com/investor-relations/governance)\\n\\t+ [Link to Earnings Call](https://www.vulcanmaterials.com/investor-relations/link-to-earnings-call)\\n\\t+ [Link to 2022 Investor Day](https://www.vulcanmaterials.com/investor-relations/link-to-2022-investor-day)\\n* [Social Responsibility](https://www.vulcanmaterials.com/social-responsibility)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/navigation-images/social-responsibility.jpg?sfvrsn=189cf41b_1)\\n\\n\\t+ [2023 Sustainability Report](https://www.vulcanmaterials.com/social-responsibility/2023-sustainability-report)\\n\\t+ [Safety \\\\& Health](https://www.vulcanmaterials.com/social-responsibility/safety-health)\\n\\t+ [Environmental Stewardship](https://www.vulcanmaterials.com/social-responsibility/environmental-stewardship)\\n\\t+ [People](https://www.vulcanmaterials.com/social-responsibility/people)\\n\\t+ [Community](https://www.vulcanmaterials.com/social-responsibility/community)\\n\\t+ [Governance](https://www.vulcanmaterials.com/social-responsibility/governance)\\n\\t+ [Teacher Center](https://www.vulcanmaterials.com/social-responsibility/teacher-center)\\n\\t+ [Vulcan Foundation](https://www.vulcanmaterials.com/social-responsibility/vulcan-foundation)\\n\\nAbout Vulcan\\n============\\n\\n* [About Vulcan](https://www.vulcanmaterials.com/about-vulcan)\\n\\n\\t+ [History](https://www.vulcanmaterials.com/about-vulcan/history)\\n\\t+ [Mission and Values](https://www.vulcanmaterials.com/about-vulcan/mission-and-values)\\n\\t+ [Executive Management](https://www.vulcanmaterials.com/about-vulcan/executive-management)\\n\\t+ [Corporate Office](https://www.vulcanmaterials.com/about-vulcan/corporate-office)\\n\\t+ [Industry Links](https://www.vulcanmaterials.com/about-vulcan/industry-links)\\n\\t+ [Careers](https://www.vulcanmaterials.com/about-vulcan/careers)\\n\\n* [Home](https://www.vulcanmaterials.com/)\\n\\\\>\\n* About Vulcan\\n\\nOur business is built on essentials. What we mine and sell—aggregates—are indispensable to infrastructure and growth. How we run our company also comes down to essentials. We run our business with great discipline and integrity. We maintain a great place to work and grow. We take care of the land through the entire cycle of productive use from mining to restoration and re\\\\-use. We serve our customers and communities well. Always with the long view in mind. This is how we sustain success.\\n\\nWe are Vulcan. And we deliver value.\\n\\n![About Vulcan](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/about-vulcan/aboutlanding_body.png?sfvrsn=c96e2a2f_1)\\n\\nVulcan Materials Company is the nation’s largest producer of construction aggregates—primarily crushed stone, sand and gravel—and a major producer of aggregates\\\\-based construction materials including asphalt and ready\\\\-mixed concrete. Our coast\\\\-to\\\\-coast footprint and strategic distribution network align with and serve the nation’s growth centers. We are headquartered in Birmingham, Alabama.\\n\\nVulcan has over 397\\\\*\\xa0active aggregates facilities, and 66\\\\*\\xa0asphalt facilities and 63\\\\*\\xa0concrete facilities, which also consume aggregates.\\xa0Most\\xa0are located in the U.S. except for our quarries and marine terminals on Vancouver Island in British Columbia, Canada,\\xa0Honduras\\xa0and Mexico’s Yucatán Peninsula.\\n\\nWhat we produce is used in nearly all forms of construction. In particular, large quantities of aggregates are used to build and repair valuable infrastructure such as roads, bridges, waterworks and ports, and to construct buildings both residential and nonresidential, such as manufacturing facilities, office buildings, schools, hospitals and churches.\\n\\n\\xa0\\\\* Data from 2023 10k and Annual Report\\n\\n[Meet Vulcan\\'s Leadership](https://ir.vulcanmaterials.com/governance/executive-management/default.aspx)\\n\\n[Vulcan Corporate Office](/about-vulcan/corporate-office)\\n\\n[Vulcan History](/about-vulcan/history)\\n\\n[Industry Links](/about-vulcan/industry-links)\\n\\n### [Contact](/contact)\\n\\n**Vulcan Materials Company**\\nCorporate Office\\n1200 Urban Center Drive\\nBirmingham, AL 35242\\nTel: (205\\\\) 298\\\\-3000\\n\\nPlease follow the hyperlinks to review the [Business Conduct Policy](https://s3.us-east-2.amazonaws.com/files.vulcanmaterials.com/vmc/Business-Conduct-Policy.pdf) or to report [concerns](https://secure.ethicspoint.com/domain/media/en/gui/16082/index.html).\\n\\nSiga los hipervínculos para revisar la\\xa0[Política de Conducta Empresarial](https://s3.us-east-2.amazonaws.com/files.vulcanmaterials.com/vmc/VMC-Business_Conduct_Policy_Spanish-2021.pdf)\\xa0o para reportar algún\\xa0[problema](https://secure.ethicspoint.com/domain/media/es/gui/16082/index.html).\\n\\nContact us for\\xa0[Product Sales](/construction-materials/product-sales).\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/footer_aboutvulcan.png?sfvrsn=a4a41105_1)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/footer_constructionmaterials.png?sfvrsn=c59fead5_1)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/footer_investorrelations.png?sfvrsn=3d684e71_1)\\n\\n![](https://vulcanmaterials.azureedge.net/content-v2/images/default-source/default-album/social-responsibility-footer.jpg?sfvrsn=c6651106_1)\\n\\nAbout Vulcan\\n\\n* [History](/about-vulcan/history)\\n* [Mission and Values](/about-vulcan/mission-and-values)\\n* [Executive Management](/about-vulcan/executive-management)\\n* [Corporate Office](/about-vulcan/corporate-office)\\n* [Industry Links](/about-vulcan/industry-links)\\n* [Careers](/about-vulcan/careers)\\n\\nConstruction Materials\\n\\n* [MyVulcan](/construction-materials/myvulcan)\\n* [Product Sales](/construction-materials/product-sales)\\n* [Products and Services](/construction-materials/products-and-services)\\n* [Suppliers](/construction-materials/suppliers)\\n* [Facilities Map](/construction-materials/facilities-map)\\n* [Operating Groups](/construction-materials/operating-groups)\\n* [Product Calculators](/construction-materials/product-calculators)\\n* [Credit Applications](/construction-materials/credit-applications)\\n* [Safety Data Sheets](/construction-materials/safety-data-sheets)\\n* [Product Declaration](/construction-materials/product-declaration)\\n* [Facilities](/construction-materials/facilities)\\n\\nInvestor Relations\\n\\n* [Overview](/investor-relations/overview)\\n* [Why Invest](/investor-relations/why-invest)\\n* [News](/investor-relations/news)\\n* [Events \\\\& Presentations](/investor-relations/events-presentations)\\n* [Stock Info](/investor-relations/stock-info)\\n* [Financials](/investor-relations/financials)\\n* [Governance](/investor-relations/governance)\\n* [Link to Earnings Call](/investor-relations/link-to-earnings-call)\\n* [Link to 2022 Investor Day](/investor-relations/link-to-2022-investor-day)\\n\\nSocial Responsibility\\n\\n* [2023 Sustainability Report](/social-responsibility/2023-sustainability-report)\\n* [Safety \\\\& Health](/social-responsibility/safety-health)\\n* [Environmental Stewardship](/social-responsibility/environmental-stewardship)\\n* [People](/social-responsibility/people)\\n* [Community](/social-responsibility/community)\\n* [Governance](/social-responsibility/governance)\\n* [Teacher Center](/social-responsibility/teacher-center)\\n* [Vulcan Foundation](/social-responsibility/vulcan-foundation)\\n\\n* [Site Map](/site-map)\\n* [Terms of Use](/terms-of-use)\\n* [Privacy Policy](/privacy-policy)\\n* [California Residents\\' Privacy Notice](/california-privacy-notice)\\n* [Contact Us About Our Site](/contact)\\n© 2024 Vulcan Materials Company', 1, 0)\n",
       "--\n",
       "tool: ('Title: Vulcan Materials Company - Wikipedia\\n=======================\\n# Vulcan Materials Company\\n\\nAmerican construction materials company\\n\\nVulcan Materials Company\\n|  | |\\n| --- | --- |\\n| Company type | Public company |\\n| [Traded as](/wiki/Ticker_symbol \"Ticker symbol\") | [NYSE](/wiki/New_York_Stock_Exchange \"New York Stock Exchange\"):\\xa0[VMC](https://www.nyse.com/quote/XNYS:VMC)[S\\\\&P 500 Component](/wiki/S%26P_500 \"S&P 500\") |\\n| Industry | [Construction materials](/wiki/Construction_materials \"Construction materials\") |\\n| Founded | 1909 as **Birmingham Slag Company** |\\n| Headquarters | [Birmingham](/wiki/Birmingham,_Alabama \"Birmingham, Alabama\"), [Alabama](/wiki/Alabama \"Alabama\") [United States](/wiki/United_States \"United States\") |\\n| Key people | [J. Thomas Hill](/wiki/J._Thomas_Hill \"J. Thomas Hill\"), [Chief Executive Officer](/wiki/Chief_Executive_Officer \"Chief Executive Officer\")Suzanne H. Wood, [Chief Financial Officer](/wiki/Chief_Financial_Officer \"Chief Financial Officer\")Denson N. Franklin III, [General Counsel](/wiki/General_Counsel \"General Counsel\") |\\n| Products | Crushed stone, sand and gravel |\\n| Revenue | Increase [$](/wiki/United_States_Dollar \"United States Dollar\")7\\\\.315 billion (2022\\\\)[\\\\[1]](#cite_note-Vulcan_2022-1) |\\n| Number of employees | 12,000[\\\\[2]](#cite_note-2)\\xa0(2023\\\\) |\\n| Website | [VulcanMaterials.com](https://www.vulcanmaterials.com) |\\n\\n**Vulcan Materials Company** ([NYSE](/wiki/New_York_Stock_Exchange \"New York Stock Exchange\"):\\xa0[VMC](https://www.nyse.com/quote/XNYS:VMC)) is an American company based in [Birmingham](/wiki/Birmingham,_Alabama \"Birmingham, Alabama\"), [Alabama](/wiki/Alabama \"Alabama\"). It is principally engaged in the production, distribution and sale of construction materials. Vulcan is the largest producer of construction materials, primarily gravel, crushed stone, and sand, and employs approximately 12,000 people at over 400 facilities. Vulcan serves 22 states, the [District of Columbia](/wiki/District_of_Columbia \"District of Columbia\"), [Mexico](/wiki/Mexico \"Mexico\"), [Canada](/wiki/Canada \"Canada\"), [Bahamas](/wiki/Bahamas \"Bahamas\") and the [U.S. Virgin Islands](/wiki/U.S._Virgin_Islands \"U.S. Virgin Islands\"). Vulcan\\'s innovative [Crescent Market project](/w/index.php?title=Crescent_Market_project&action=edit&redlink=1 \"Crescent Market project (page does not exist)\") led to construction of a large quarry and deep water seaport on the [Yucatán Peninsula](/wiki/Yucat%C3%A1n_Peninsula \"Yucatán Peninsula\") of Mexico, just south of [Cancun](/wiki/Cancun \"Cancun\"). This quarry supplies Tampa, New Orleans, Houston, and Brownsville, Texas, as well as other Gulf coast seaports, with crushed limestone via large 62,000\\\\-ton self\\\\-discharging ships.\\n\\nHistory\\n-------\\n\\n\\\\[[edit](/w/index.php?title=Vulcan_Materials_Company&action=edit&section=1 \"Edit section: History\")]\\nIn the late 1980s Vulcan acquired Reed Crushed Stone, near [Kentucky](/wiki/Gravel_Switch,_Livingston_County,_Kentucky \"Gravel Switch, Livingston County, Kentucky\").[\\\\[3]](#cite_note-3)[\\\\[4]](#cite_note-4)\\n\\nOn November 16, 1998, Vulcan announced it was purchasing CalMat Company, a producer of asphalt and ready\\\\-mixed concrete based in [Los Angeles](/wiki/Los_Angeles \"Los Angeles\"), for $760 million in cash.[\\\\[5]](#cite_note-5)\\n\\nOn June 7, 2005, Vulcan completed the sale of its chemicals business, known as Vulcan Chemicals, to [Occidental Chemical Corporation](/wiki/Occidental_Chemical_Corporation \"Occidental Chemical Corporation\"). The sale of assets included Vulcan\\'s [chloralkali](/wiki/Chloralkali_process \"Chloralkali process\") plants in Wichita, Kansas; Geismar, Louisiana; and Port Edwards, Wisconsin, as well as Vulcan Chemicals\\' joint venture located in Geismar, Louisiana. In 2005, Vulcan acquired 11 aggregates operations and five asphalt plants in Arizona, Georgia, Indiana and Tennessee.\\n\\nOn February 19, 2007, Vulcan announced that it would buy stone and cement producer Florida Rock Industries for $4\\\\.7 billion.[\\\\[6]](#cite_note-6)[\\\\[7]](#cite_note-7) Vulcan completed the acquisition of Florida Rock on November 16, 2007\\\\.[\\\\[8]](#cite_note-8)\\n\\nIn March 2007, Vulcan announced that it had been named to [Fortune Magazine](/wiki/Fortune_Magazine \"Fortune Magazine\")\\'s list of *Most Admired Companies* for the sixth time. The company was ranked first in its industry sector, \"Building Materials, Glass.\" Overall, Vulcan ranked among the top 10 companies in the [Fortune 1000](/wiki/Fortune_1000 \"Fortune 1000\") for both long\\\\-term investment and [social responsibility](/wiki/Social_responsibility \"Social responsibility\").\\n\\nIn January 2014, Vulcan Materials announced they would be selling their Florida Rock Industry (cement and ready\\\\-mixed concrete) plants and equipment to [Cementos Argos](/wiki/Argos_S.A. \"Argos S.A.\"), of Colombia. In March 2014, Vulcan was able to fully divest these properties to Cementos Argos.[\\\\[9]](#cite_note-FloridaRockSale-9)\\n\\nIn January 2018, Vulcan Materials was found to be partially responsible for the [Bayou Corne sinkhole](/wiki/Bayou_Corne_sinkhole \"Bayou Corne sinkhole\"), along with [Occidental Petroleum](/wiki/Occidental_Petroleum \"Occidental Petroleum\") and Texas Brine Company.[\\\\[10]](#cite_note-10)\\n\\nIn June 2021, Vulcan Materials announced they would be acquiring [US Concrete (USCR)](/wiki/U.S._Concrete \"U.S. Concrete\") for $74 per share, a 30% premium on the date of the announcement. U.S. Concrete was formed in 1999, with its headquarters in Euless, TX. The deal is valued at $1\\\\.3 billion. The acquisition will allow for integrated expansion of Vulcan and make the subsidiary debt\\\\-free with adequate working capital. The deal has been approved by both boards and will close in the second half of 2021\\\\.[\\\\[11]](#cite_note-USConcrete_Acquisition-11)\\n\\nIn early 2023 after being accused by president [Andrés Manuel López Obrador](/wiki/L%C3%B3pez_Obrador \"López Obrador\") of exporting materials from Mexico without a permit, the Mexican military seized a marine terminal belonging to Vulcan leading to a diplomatic spat between the US and Mexico.[\\\\[12]](#cite_note-12)\\n\\nReferences\\n----------\\n\\n\\\\[[edit](/w/index.php?title=Vulcan_Materials_Company&action=edit&section=2 \"Edit section: References\")]\\n\\n1. **[^](#cite_ref-Vulcan_2022_1-0)** [\"VULCAN REPORTS FOURTH QUARTER AND FULL YEAR 2022 RESULTS\"](https://ir.vulcanmaterials.com/news/news-details/2023/VULCAN-REPORTS-FOURTH-QUARTER-AND-FULL-YEAR-2022-RESULTS/default.aspx). Vulcan Materials Company. Retrieved Apr 2, 2023.\\n2. **[^](#cite_ref-2)** [\"Vulcan Materials\"](https://www.sec.gov/ix?doc=/Archives/edgar/data/1396009/000139600923000007/vmc-20221231x10k.htm#PartII_Item8). *SEC.gov*. Retrieved 2023\\\\-04\\\\-02.\\n3. **[^](#cite_ref-3)** [*Minerals Yearbook*](https://books.google.com/books?id=WkoZSm8sXQIC&q=Reed+Terminal+aggregates&pg=PA211). 1989\\\\. [ISBN](/wiki/ISBN_(identifier) \"ISBN (identifier)\")\\xa0[9780160359316](/wiki/Special:BookSources/9780160359316 \"Special:BookSources/9780160359316\").\\n4. **[^](#cite_ref-4)** [*Minerals Yearbook*](https://books.google.com/books?id=EDW2hvyndDUC&q=vulcan+mineral+grand+river+operations&pg=PA191). 2009\\\\. [ISBN](/wiki/ISBN_(identifier) \"ISBN (identifier)\")\\xa0[9781411323292](/wiki/Special:BookSources/9781411323292 \"Special:BookSources/9781411323292\").\\n5. **[^](#cite_ref-5)** Vrana, Debora (November 16, 1998\\\\). [\"Concrete Firm CalMat to Be Acquired for $760 Million\"](https://www.latimes.com/archives/la-xpm-1998-nov-16-fi-43297-story.html). *The Los Angeles Times*. Retrieved 2024\\\\-08\\\\-10.\\n6. **[^](#cite_ref-6)** Dolmetsch, Chris \\\\& Lococo, Edmond (19 Feb 2007\\\\). [\"Vulcan Materials to Buy Florida Rock for $4\\\\.6 Billion (Update4\\\\)\"](https://www.bloomberg.com/apps/news?pid=newsarchive&sid=amqbTdNtr8Pw). *Bloomberg*. Retrieved 12 May 2015.\\n7. **[^](#cite_ref-7)** [\"Vulcan Materials to buy Florida Rock for $4\\\\.6 billion\"](https://www.reuters.com/article/us-floridarock-vulcan-idUSN1944560620070220). *[Reuters](/wiki/Reuters \"Reuters\")*. 19 February 2007. Retrieved 27 May 2021.\\n8. **[^](#cite_ref-8)** Zubko, Nick (19 March 2008\\\\). [\"Florida Rock Industries Inc.: Under New Management\"](https://www.industryweek.com/leadership/companies-executives/article/21942309/florida-rock-industries-inc-under-new-management). *Industry Week*. Retrieved 27 May 2021.\\n9. **[^](#cite_ref-FloridaRockSale_9-0)** Basch, Mark (27 January 2014\\\\). [\"Vulcan selling former Florida Rock cement facilities\"](https://www.jaxdailyrecord.com/showstory.php?Story_id=541695). *Jacksonville Daily Record*. Retrieved 12 May 2015.\\n10. **[^](#cite_ref-10)** [Judge: Fault for Bayou Corne sinkhole lies with Texas Brine, OxyChem, Vulcan; companies had decades of warnings](https://www.theadvocate.com/baton_rouge/news/article_7856ff5e-f4ae-11e7-b86c-4f261682612b.html) [The Advocate](/wiki/The_Advocate_(Louisiana) \"The Advocate (Louisiana)\")\\n11. **[^](#cite_ref-USConcrete_Acquisition_11-0)** Lenihan, Rob (7 June 2021\\\\). [\"U.S. Concrete Skyrockets on Deal to Be Bought for $1\\\\.29B by Vulcan\"](https://www.thestreet.com/investing/us-concrete-uscr-vulcan-materials-vmc-acquisition). *The Street*. Retrieved 7 Jun 2021.\\n12. **[^](#cite_ref-12)** Haldevang, Max (19 Mar 2023\\\\). [\"Vulcan Facility Seizure Adds to Tension Between US and Mexico\"](https://www.bloomberg.com/news/articles/2023-03-19/vulcan-facility-seizure-adds-to-tension-between-us-and-mexico?leadSource=uverify%20wall).\\n\\nExternal links\\n--------------\\n\\n\\\\[[edit](/w/index.php?title=Vulcan_Materials_Company&action=edit&section=3 \"Edit section: External links\")]\\n* [Vulcan Materials website](http://www.vulcanmaterials.com)\\n* [*U.S. v. Vulcan Materials Co. and Florida Rock Industries Inc.*](https://www.justice.gov/atr/case/us-v-vulcan-materials-co-and-florida-rock-industries-inc), a US Supreme court case\\n* Business data for Vulcan Materials Company:\\n\\t+ [Google](https://www.google.com/finance/quote/VMC)\\n\\t+ [SEC filings](https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=VMC)\\n\\t+ [Yahoo!](https://finance.yahoo.com/quote/VMC)\\n\\n| [Authority control databases](/wiki/Help:Authority_control \"Help:Authority control\") [Edit this at Wikidata](https://www.wikidata.org/wiki/Q1685287#identifiers \"Edit this at Wikidata\") | |\\n| --- | --- |\\n| International | * [ISNI](https://isni.org/isni/0000000449084962) * [VIAF](https://viaf.org/viaf/131972586) |\\n| National | * [United States](https://id.loc.gov/authorities/n84041729) |\\n\\n![](https://login.wikimedia.org/wiki/Special:CentralAutoLogin/start?useformat=desktop&type=1x1&usesul3=0)\\nRetrieved from \"[https://en.wikipedia.org/w/index.php?title\\\\=Vulcan\\\\_Materials\\\\_Company\\\\&oldid\\\\=1248703031](https://en.wikipedia.org/w/index.php?title=Vulcan_Materials_Company&oldid=1248703031)\"', 1, 0)</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 484 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=\"Vulcan Materials Company is a prominent American company specializing in the production, distribution, and sale of construction materials. Here's an overview based on the information from their official site and Wikipedia:\\n\\n### Overview\\n- **Industry**: Construction materials\\n- **Founded**: 1909 as Birmingham Slag Company\\n- **Headquarters**: Birmingham, Alabama, United States\\n- **Key People**: J. Thomas Hill (CEO), Suzanne H. Wood (CFO), Denson N. Franklin III (General Counsel)\\n- **Products**: Crushed stone, sand, gravel, asphalt, and ready-mixed concrete\\n- **Revenue**: $7.315 billion (2022)\\n- **Employees**: Approximately 12,000 (2023)\\n- **Website**: [VulcanMaterials.com](https://www.vulcanmaterials.com)\\n\\n### Operations\\nVulcan Materials is the largest producer of construction aggregates in the U.S., including crushed stone, sand, and gravel. The company operates over 400 facilities across 22 states, the District of Columbia, Mexico, Canada, the Bahamas, and the U.S. Virgin Islands. They have a significant presence in the Gulf Coast with a quarry and deep-water seaport on the Yucatán Peninsula in Mexico.\\n\\n### History\\n- **1980s**: Acquired Reed Crushed Stone in Kentucky.\\n- **1998**: Purchased CalMat Company for $760 million.\\n- **2005**: Sold its chemicals business to Occidental Chemical Corporation.\\n- **2007**: Acquired Florida Rock Industries for $4.7 billion.\\n- **2014**: Sold Florida Rock Industry plants to Cementos Argos.\\n- **2021**: Announced acquisition of U.S. Concrete for $1.3 billion.\\n\\n### Recent Developments\\nIn early 2023, Vulcan faced a diplomatic issue when the Mexican military seized one of its marine terminals, leading to tensions between the U.S. and Mexico.\\n\\n### Social Responsibility\\nVulcan emphasizes sustainability, safety, health, environmental stewardship, and community engagement. They publish an annual sustainability report and have been recognized for their social responsibility efforts.\\n\\nFor more detailed information, you can visit their [official website](https://www.vulcanmaterials.com) or their [Wikipedia page](https://en.wikipedia.org/wiki/Vulcan_Materials_Company).\", role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box respond'><h4 class='step-header'>[14] search_agent responds to analyst</h4><pre class='step-text'>kind: respond\n",
       "copy_output: false\n",
       "\n",
       "<details><summary>2125 characters ...</summary>Vulcan Materials Company is a prominent American company specializing in the production, distribution, and sale of construction materials. Here's an overview based on the information from their official site and Wikipedia:\n",
       "\n",
       "### Overview\n",
       "- **Industry**: Construction materials\n",
       "- **Founded**: 1909 as Birmingham Slag Company\n",
       "- **Headquarters**: Birmingham, Alabama, United States\n",
       "- **Key People**: J. Thomas Hill (CEO), Suzanne H. Wood (CFO), Denson N. Franklin III (General Counsel)\n",
       "- **Products**: Crushed stone, sand, gravel, asphalt, and ready-mixed concrete\n",
       "- **Revenue**: $7.315 billion (2022)\n",
       "- **Employees**: Approximately 12,000 (2023)\n",
       "- **Website**: [VulcanMaterials.com](https://www.vulcanmaterials.com)\n",
       "\n",
       "### Operations\n",
       "Vulcan Materials is the largest producer of construction aggregates in the U.S., including crushed stone, sand, and gravel. The company operates over 400 facilities across 22 states, the District of Columbia, Mexico, Canada, the Bahamas, and the U.S. Virgin Islands. They have a significant presence in the Gulf Coast with a quarry and deep-water seaport on the Yucatán Peninsula in Mexico.\n",
       "\n",
       "### History\n",
       "- **1980s**: Acquired Reed Crushed Stone in Kentucky.\n",
       "- **1998**: Purchased CalMat Company for $760 million.\n",
       "- **2005**: Sold its chemicals business to Occidental Chemical Corporation.\n",
       "- **2007**: Acquired Florida Rock Industries for $4.7 billion.\n",
       "- **2014**: Sold Florida Rock Industry plants to Cementos Argos.\n",
       "- **2021**: Announced acquisition of U.S. Concrete for $1.3 billion.\n",
       "\n",
       "### Recent Developments\n",
       "In early 2023, Vulcan faced a diplomatic issue when the Mexican military seized one of its marine terminals, leading to tensions between the U.S. and Mexico.\n",
       "\n",
       "### Social Responsibility\n",
       "Vulcan emphasizes sustainability, safety, health, environmental stewardship, and community engagement. They publish an annual sustainability report and have been recognized for their social responsibility efforts.\n",
       "\n",
       "For more detailed information, you can visit their [official website](https://www.vulcanmaterials.com) or their [Wikipedia page](https://en.wikipedia.org/wiki/Vulcan_Materials_Company).</details></pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 705 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_ticker\",\n",
       "      \"description\": \"Get company stock ticker from its name.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"company_name\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"company_name\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_data\",\n",
       "      \"description\": \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"symbol\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"start_date\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"end_date\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"symbol\",\n",
       "          \"start_date\",\n",
       "          \"end_date\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"call_search_agent\",\n",
       "      \"description\": \"Use this tool to ask a fellow AI agent to search for information on the web.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "system: \n",
       "You will help the user to learn about financials of companies. \n",
       "For general user queries, include some info about stock price changes during the last year, as well as some general information on the company.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan.\n",
       "2. Use the `functions.call_search_agent` tool to gather general information about Vulcan, including its business operations and industry.\n",
       "3. Use the `functions.get_stock_data` tool to retrieve Vulcan's stock price data for the past year, using the stock ticker obtained in step 1.\n",
       "4. Compile the information from steps 2 and 3 into a concise summary about Vulcan, including stock price changes over the last year and general company information.\n",
       "--\n",
       "assistant: [{'function': {'name': 'get_stock_ticker', 'arguments': '{\"company_name\":\"Vulcan\"}'}, 'id': 'call_MnfVd4XRZFgrixiBPT7QAvrn', 'type': 'function'}]\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "assistant: Call search_agent with the message 'Vulcan company overview'\n",
       "--\n",
       "user (search_agent): Vulcan Materials Company is a prominent American company specializing in the production, distribution, and sale of construction materials. Here's an overview based on the information from their official site and Wikipedia:\n",
       "\n",
       "### Overview\n",
       "- **Industry**: Construction materials\n",
       "- **Founded**: 1909 as Birmingham Slag Company\n",
       "- **Headquarters**: Birmingham, Alabama, United States\n",
       "- **Key People**: J. Thomas Hill (CEO), Suzanne H. Wood (CFO), Denson N. Franklin III (General Counsel)\n",
       "- **Products**: Crushed stone, sand, gravel, asphalt, and ready-mixed concrete\n",
       "- **Revenue**: $7.315 billion (2022)\n",
       "- **Employees**: Approximately 12,000 (2023)\n",
       "- **Website**: [VulcanMaterials.com](https://www.vulcanmaterials.com)\n",
       "\n",
       "### Operations\n",
       "Vulcan Materials is the largest producer of construction aggregates in the U.S., including crushed stone, sand, and gravel. The company operates over 400 facilities across 22 states, the District of Columbia, Mexico, Canada, the Bahamas, and the U.S. Virgin Islands. They have a significant presence in the Gulf Coast with a quarry and deep-water seaport on the Yucatán Peninsula in Mexico.\n",
       "\n",
       "### History\n",
       "- **1980s**: Acquired Reed Crushed Stone in Kentucky.\n",
       "- **1998**: Purchased CalMat Company for $760 million.\n",
       "- **2005**: Sold its chemicals business to Occidental Chemical Corporation.\n",
       "- **2007**: Acquired Florida Rock Industries for $4.7 billion.\n",
       "- **2014**: Sold Florida Rock Industry plants to Cementos Argos.\n",
       "- **2021**: Announced acquisition of U.S. Concrete for $1.3 billion.\n",
       "\n",
       "### Recent Developments\n",
       "In early 2023, Vulcan faced a diplomatic issue when the Mexican military seized one of its marine terminals, leading to tensions between the U.S. and Mexico.\n",
       "\n",
       "### Social Responsibility\n",
       "Vulcan emphasizes sustainability, safety, health, environmental stewardship, and community engagement. They publish an annual sustainability report and have been recognized for their social responsibility efforts.\n",
       "\n",
       "For more detailed information, you can visit their [official website](https://www.vulcanmaterials.com) or their [Wikipedia page](https://en.wikipedia.org/wiki/Vulcan_Materials_Company).\n",
       "--\n",
       "user: Follow the plan you created to earlier. When you are done, respond to the user.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 0 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=None, role='assistant', tool_calls=[ChatCompletionMessageToolCall(function=Function(arguments='{\"symbol\":\"VMC\",\"start_date\":\"2023-09-17\",\"end_date\":\"2024-09-17\"}', name='get_stock_data'), id='call_0iD05OgtIBscmvfeSiU0ZjQg', type='function')], function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box thought'><h4 class='step-header'>[15] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: act\n",
       "</pre></div><div class='basic-renderer-box action'><h4 class='step-header'>[16] Action: ToolCalls</h4><pre class='step-text'>kind: assistant\n",
       "tool_calls:\n",
       "- function:\n",
       "    name: get_stock_data\n",
       "    arguments: '{\"symbol\":\"VMC\",\"start_date\":\"2023-09-17\",\"end_date\":\"2024-09-17\"}'\n",
       "  id: call_0iD05OgtIBscmvfeSiU0ZjQg\n",
       "  type: function\n",
       "</pre></div><div class='basic-renderer-box observation'><h4 class='step-header'>[17] Observation: ToolResult</h4><pre class='step-text'>kind: tool\n",
       "tool_call_id: call_0iD05OgtIBscmvfeSiU0ZjQg\n",
       "\n",
       "<details><summary>2195 characters ...</summary>[['2023-09-18', 211.69000244140625], ['2023-09-22', 200.6199951171875], ['2023-09-28', 205.02999877929688], ['2023-10-04', 205.02999877929688], ['2023-10-10', 211.0], ['2023-10-16', 213.0399932861328], ['2023-10-20', 201.5500030517578], ['2023-10-26', 193.97000122070312], ['2023-11-01', 203.8000030517578], ['2023-11-07', 207.2100067138672], ['2023-11-13', 210.49000549316406], ['2023-11-17', 212.35000610351562], ['2023-11-24', 211.67999267578125], ['2023-11-30', 213.55999755859375], ['2023-12-06', 211.92999267578125], ['2023-12-12', 220.89999389648438], ['2023-12-18', 222.6999969482422], ['2023-12-22', 224.92999267578125], ['2023-12-29', 227.00999450683594], ['2024-01-05', 221.6199951171875], ['2024-01-11', 224.36000061035156], ['2024-01-18', 225.1199951171875], ['2024-01-24', 219.8000030517578], ['2024-01-30', 231.0500030517578], ['2024-02-05', 229.64999389648438], ['2024-02-09', 240.0], ['2024-02-15', 242.4600067138672], ['2024-02-22', 256.94000244140625], ['2024-02-28', 262.29998779296875], ['2024-03-05', 264.9800109863281], ['2024-03-11', 264.95001220703125], ['2024-03-15', 266.8599853515625], ['2024-03-21', 275.5899963378906], ['2024-03-27', 272.7900085449219], ['2024-04-03', 268.7699890136719], ['2024-04-09', 265.6199951171875], ['2024-04-15', 260.4700012207031], ['2024-04-19', 252.05999755859375], ['2024-04-25', 258.5400085449219], ['2024-05-01', 259.7300109863281], ['2024-05-07', 268.3500061035156], ['2024-05-13', 270.0899963378906], ['2024-05-17', 259.1000061035156], ['2024-05-23', 257.2099914550781], ['2024-05-30', 254.0800018310547], ['2024-06-05', 251.17999267578125], ['2024-06-11', 244.63999938964844], ['2024-06-17', 256.3500061035156], ['2024-06-24', 247.5800018310547], ['2024-06-28', 248.67999267578125], ['2024-07-05', 244.2100067138672], ['2024-07-11', 254.69000244140625], ['2024-07-17', 255.8000030517578], ['2024-07-23', 260.1499938964844], ['2024-07-29', 264.5299987792969], ['2024-08-02', 266.3399963378906], ['2024-08-08', 246.3300018310547], ['2024-08-14', 242.5], ['2024-08-20', 245.60000610351562], ['2024-08-26', 250.67999267578125], ['2024-08-30', 245.2100067138672], ['2024-09-06', 231.97000122070312], ['2024-09-12', 236.27000427246094]]</details></pre></div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "        <details>\n",
       "            <summary>Prompt 1790 tokens </summary>\n",
       "            <pre style='font-size: 12px; white-space: pre-wrap;word-wrap: break-word;'>tool_schemas: [\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_ticker\",\n",
       "      \"description\": \"Get company stock ticker from its name.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"company_name\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"company_name\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"get_stock_data\",\n",
       "      \"description\": \"Get stock proces for a given symbol and date range.\\n\\n    Args:\\n        symbol (str): Stock symbol.\\n        start_date (str): Start date in the format 'YYYY-MM-DD'.\\n        end_date (str): End date in the format 'YYYY-MM-DD'.\\n\\n    Returns:\\n        (list[tuple]): List of tuples, each tuple contains a 'YYYY-MM-DD' date and the stock price.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"symbol\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"start_date\": {\n",
       "            \"type\": \"string\"\n",
       "          },\n",
       "          \"end_date\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"symbol\",\n",
       "          \"start_date\",\n",
       "          \"end_date\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  },\n",
       "  {\n",
       "    \"type\": \"function\",\n",
       "    \"function\": {\n",
       "      \"name\": \"call_search_agent\",\n",
       "      \"description\": \"Use this tool to ask a fellow AI agent to search for information on the web.\",\n",
       "      \"parameters\": {\n",
       "        \"properties\": {\n",
       "          \"query\": {\n",
       "            \"type\": \"string\"\n",
       "          }\n",
       "        },\n",
       "        \"required\": [\n",
       "          \"query\"\n",
       "        ],\n",
       "        \"type\": \"object\"\n",
       "      }\n",
       "    }\n",
       "  }\n",
       "]\n",
       "--\n",
       "system: \n",
       "You will help the user to learn about financials of companies. \n",
       "For general user queries, include some info about stock price changes during the last year, as well as some general information on the company.\n",
       "Today is 2024-09-17.\n",
       "\n",
       "--\n",
       "user: Tell me about Vulcan in 3 sentences\n",
       "--\n",
       "assistant: 1. Use the `functions.get_stock_ticker` tool to find the stock ticker symbol for Vulcan.\n",
       "2. Use the `functions.call_search_agent` tool to gather general information about Vulcan, including its business operations and industry.\n",
       "3. Use the `functions.get_stock_data` tool to retrieve Vulcan's stock price data for the past year, using the stock ticker obtained in step 1.\n",
       "4. Compile the information from steps 2 and 3 into a concise summary about Vulcan, including stock price changes over the last year and general company information.\n",
       "--\n",
       "assistant: [{'function': {'name': 'get_stock_ticker', 'arguments': '{\"company_name\":\"Vulcan\"}'}, 'id': 'call_MnfVd4XRZFgrixiBPT7QAvrn', 'type': 'function'}]\n",
       "--\n",
       "tool: VMC\n",
       "--\n",
       "assistant: Call search_agent with the message 'Vulcan company overview'\n",
       "--\n",
       "user (search_agent): Vulcan Materials Company is a prominent American company specializing in the production, distribution, and sale of construction materials. Here's an overview based on the information from their official site and Wikipedia:\n",
       "\n",
       "### Overview\n",
       "- **Industry**: Construction materials\n",
       "- **Founded**: 1909 as Birmingham Slag Company\n",
       "- **Headquarters**: Birmingham, Alabama, United States\n",
       "- **Key People**: J. Thomas Hill (CEO), Suzanne H. Wood (CFO), Denson N. Franklin III (General Counsel)\n",
       "- **Products**: Crushed stone, sand, gravel, asphalt, and ready-mixed concrete\n",
       "- **Revenue**: $7.315 billion (2022)\n",
       "- **Employees**: Approximately 12,000 (2023)\n",
       "- **Website**: [VulcanMaterials.com](https://www.vulcanmaterials.com)\n",
       "\n",
       "### Operations\n",
       "Vulcan Materials is the largest producer of construction aggregates in the U.S., including crushed stone, sand, and gravel. The company operates over 400 facilities across 22 states, the District of Columbia, Mexico, Canada, the Bahamas, and the U.S. Virgin Islands. They have a significant presence in the Gulf Coast with a quarry and deep-water seaport on the Yucatán Peninsula in Mexico.\n",
       "\n",
       "### History\n",
       "- **1980s**: Acquired Reed Crushed Stone in Kentucky.\n",
       "- **1998**: Purchased CalMat Company for $760 million.\n",
       "- **2005**: Sold its chemicals business to Occidental Chemical Corporation.\n",
       "- **2007**: Acquired Florida Rock Industries for $4.7 billion.\n",
       "- **2014**: Sold Florida Rock Industry plants to Cementos Argos.\n",
       "- **2021**: Announced acquisition of U.S. Concrete for $1.3 billion.\n",
       "\n",
       "### Recent Developments\n",
       "In early 2023, Vulcan faced a diplomatic issue when the Mexican military seized one of its marine terminals, leading to tensions between the U.S. and Mexico.\n",
       "\n",
       "### Social Responsibility\n",
       "Vulcan emphasizes sustainability, safety, health, environmental stewardship, and community engagement. They publish an annual sustainability report and have been recognized for their social responsibility efforts.\n",
       "\n",
       "For more detailed information, you can visit their [official website](https://www.vulcanmaterials.com) or their [Wikipedia page](https://en.wikipedia.org/wiki/Vulcan_Materials_Company).\n",
       "--\n",
       "assistant: [{'function': {'name': 'get_stock_data', 'arguments': '{\"symbol\":\"VMC\",\"start_date\":\"2023-09-17\",\"end_date\":\"2024-09-17\"}'}, 'id': 'call_0iD05OgtIBscmvfeSiU0ZjQg', 'type': 'function'}]\n",
       "--\n",
       "tool: [['2023-09-18', 211.69000244140625], ['2023-09-22', 200.6199951171875], ['2023-09-28', 205.02999877929688], ['2023-10-04', 205.02999877929688], ['2023-10-10', 211.0], ['2023-10-16', 213.0399932861328], ['2023-10-20', 201.5500030517578], ['2023-10-26', 193.97000122070312], ['2023-11-01', 203.8000030517578], ['2023-11-07', 207.2100067138672], ['2023-11-13', 210.49000549316406], ['2023-11-17', 212.35000610351562], ['2023-11-24', 211.67999267578125], ['2023-11-30', 213.55999755859375], ['2023-12-06', 211.92999267578125], ['2023-12-12', 220.89999389648438], ['2023-12-18', 222.6999969482422], ['2023-12-22', 224.92999267578125], ['2023-12-29', 227.00999450683594], ['2024-01-05', 221.6199951171875], ['2024-01-11', 224.36000061035156], ['2024-01-18', 225.1199951171875], ['2024-01-24', 219.8000030517578], ['2024-01-30', 231.0500030517578], ['2024-02-05', 229.64999389648438], ['2024-02-09', 240.0], ['2024-02-15', 242.4600067138672], ['2024-02-22', 256.94000244140625], ['2024-02-28', 262.29998779296875], ['2024-03-05', 264.9800109863281], ['2024-03-11', 264.95001220703125], ['2024-03-15', 266.8599853515625], ['2024-03-21', 275.5899963378906], ['2024-03-27', 272.7900085449219], ['2024-04-03', 268.7699890136719], ['2024-04-09', 265.6199951171875], ['2024-04-15', 260.4700012207031], ['2024-04-19', 252.05999755859375], ['2024-04-25', 258.5400085449219], ['2024-05-01', 259.7300109863281], ['2024-05-07', 268.3500061035156], ['2024-05-13', 270.0899963378906], ['2024-05-17', 259.1000061035156], ['2024-05-23', 257.2099914550781], ['2024-05-30', 254.0800018310547], ['2024-06-05', 251.17999267578125], ['2024-06-11', 244.63999938964844], ['2024-06-17', 256.3500061035156], ['2024-06-24', 247.5800018310547], ['2024-06-28', 248.67999267578125], ['2024-07-05', 244.2100067138672], ['2024-07-11', 254.69000244140625], ['2024-07-17', 255.8000030517578], ['2024-07-23', 260.1499938964844], ['2024-07-29', 264.5299987792969], ['2024-08-02', 266.3399963378906], ['2024-08-08', 246.3300018310547], ['2024-08-14', 242.5], ['2024-08-20', 245.60000610351562], ['2024-08-26', 250.67999267578125], ['2024-08-30', 245.2100067138672], ['2024-09-06', 231.97000122070312], ['2024-09-12', 236.27000427246094]]\n",
       "--\n",
       "user: Follow the plan you created to earlier. When you are done, respond to the user.</pre>\n",
       "        </details>\n",
       "        </div><div class='basic-prompt-box' style='background-color:#ffffba; margin: 0 4px;'>\n",
       "                <details>\n",
       "                    <summary>LLM Output 129 tokens</summary>\n",
       "                    <pre style='font-size: 12px; white-space: pre-wrap; word-wrap: break-word;'>Message(content=\"Vulcan Materials Company, founded in 1909 and headquartered in Birmingham, Alabama, is the largest producer of construction aggregates in the U.S., specializing in crushed stone, sand, gravel, asphalt, and ready-mixed concrete. Over the past year, Vulcan's stock (ticker symbol: VMC) has experienced fluctuations, starting at approximately $211.69 in September 2023, peaking at around $275.59 in March 2024, and recently trading at about $236.27 as of September 2024. The company operates over 400 facilities across multiple countries and emphasizes sustainability and community engagement in its operations.\", role='assistant', tool_calls=None, function_call=None, provider_specific_fields=None, refusal=None)</pre>\n",
       "                </details>\n",
       "                </div><div class='basic-renderer-box thought'><h4 class='step-header'>[18] Thought: SetNextNode</h4><pre class='step-text'>kind: set_next_node\n",
       "next_node: plan\n",
       "</pre></div><div class='basic-renderer-box action'><h4 class='step-header'>[19] Assistant</h4><pre class='step-text'>kind: assistant\n",
       "\n",
       "Vulcan Materials Company, founded in 1909 and headquartered in Birmingham, Alabama, is the largest producer of construction aggregates in the U.S., specializing in crushed stone, sand, gravel, asphalt, and ready-mixed concrete. Over the past year, Vulcan's stock (ticker symbol: VMC) has experienced fluctuations, starting at approximately $211.69 in September 2023, peaking at around $275.59 in March 2024, and recently trading at about $236.27 as of September 2024. The company operates over 400 facilities across multiple countries and emphasizes sustainability and community engagement in its operations.</pre></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from IPython.display import clear_output\n",
    "\n",
    "from tapeagents.core import SetNextNode\n",
    "from tapeagents.dialog_tape import AssistantThought, ToolCalls\n",
    "from tapeagents.orchestrator import MainLoopStatus, main_loop\n",
    "from tapeagents.view import Call\n",
    "\n",
    "system_instruction = f\"\"\"\n",
    "You will help the user to learn about financials of companies. \n",
    "For general user queries, include some info about stock price changes during the last year, as well as some general information on the company.\n",
    "Today is {today}.\n",
    "\"\"\"\n",
    "system_message = {\"role\": \"system\", \"content\": system_instruction}\n",
    "\n",
    "\n",
    "class PlanNode(Node):\n",
    "    name: str = \"plan\"\n",
    "\n",
    "    def make_prompt(self, agent, tape) -> Prompt:\n",
    "        view = agent.compute_view(tape)\n",
    "        guidance = \"Write a natural language plan on how to use tools help the user. Output a list of numbered items, like 1., 2., 3., etc.\"\n",
    "        guidance_message = {\"role\": \"user\", \"content\": guidance}\n",
    "        return Prompt(\n",
    "            messages=[system_message] + view_to_messages(view.top, agent) + [guidance_message],\n",
    "            tools=main_agent_env.get_tool_schema_dicts(),\n",
    "        )\n",
    "\n",
    "    def generate_steps(self, agent, dialog, llm_stream: LLMStream):\n",
    "        if content := llm_stream.get_output().content:\n",
    "            yield AssistantThought(content=content)\n",
    "        else:\n",
    "            raise ValueError()\n",
    "\n",
    "\n",
    "class ActNode(Node):\n",
    "    name: str = \"act\"\n",
    "\n",
    "    def make_prompt(self, agent, tape: DialogTape) -> Prompt:\n",
    "        view = agent.compute_view(tape)\n",
    "        guidance = \"Follow the plan you created to earlier. When you are done, respond to the user.\"\n",
    "        guidance_message = {\"role\": \"user\", \"content\": guidance}\n",
    "        return Prompt(\n",
    "            messages=[system_message] + view_to_messages(view.top, agent) + [guidance_message],\n",
    "            tools=main_agent_env.get_tool_schema_dicts(),\n",
    "        )\n",
    "\n",
    "    def generate_steps(self, agent, dialog, llm_stream: LLMStream):\n",
    "        o = llm_stream.get_output()\n",
    "        if o.content:\n",
    "            yield SetNextNode(next_node=\"plan\")\n",
    "            yield AssistantStep(content=o.content)\n",
    "        elif o.tool_calls:\n",
    "            yield SetNextNode(next_node=\"act\")\n",
    "            # only keep the tool calls before the call to another agent\n",
    "            agent_call = None\n",
    "            for i, tc in enumerate(o.tool_calls):\n",
    "                if tc.function.name == \"call_search_agent\":\n",
    "                    agent_call = tc\n",
    "                    o.tool_calls = o.tool_calls[:i]\n",
    "                    break\n",
    "            # either produce the ToolCalls action OR call another agent\n",
    "            if o.tool_calls:\n",
    "                yield ToolCalls.from_llm_output(o)\n",
    "            else:\n",
    "                assert agent_call and agent_call.function.name == \"call_search_agent\"\n",
    "                yield Call(agent_name=\"search_agent\", content=json.loads(agent_call.function.arguments)[\"query\"])\n",
    "\n",
    "        else:\n",
    "            raise ValueError()\n",
    "\n",
    "\n",
    "multi_agent_analyst = Agent.create(\n",
    "    name=\"analyst\",\n",
    "    subagents=[search_agent.clone()],\n",
    "    llms=LiteLLM(model_name=\"gpt-4o\", parameters={\"temperature\": 0.1}),\n",
    "    nodes=[PlanNode(), ActNode()],\n",
    ")\n",
    "\n",
    "print(\"Run the agent!\")\n",
    "start_tape = DialogTape(steps=[UserStep(content=\"Tell me about Vulcan in 3 sentences\")])\n",
    "for event in main_loop(multi_agent_analyst, start_tape, whole_env):\n",
    "    # This agent runs for a while, so we will show you a fresh render every time\n",
    "    # when the environment finishes reacting with new actions\n",
    "    if new_tape := event.agent_tape or event.env_tape:\n",
    "        clear_output()\n",
    "        display(HTML(render_tape_with_prompts(new_tape, PrettyRenderer())))\n",
    "    # Uncomment this if you want to pause after every loop\n",
    "    # if event.env_tape:\n",
    "    #     input(\"Press Enter the run the next iteration of the main loop\")\n",
    "    if event.status == MainLoopStatus.EXTERNAL_INPUT_NEEDED:\n",
    "        break"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Look at this rather long tape and note how \"analyst\" calls \"search_agent\", and how the latter then responds to \"analyst\". Congratulations, you now know how to build a multi-agent TapeAgent!"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tapeagents",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}