Bedrock sample (#116)

* init + fix: logging

* workflow 2 now gives a chat summary at the end. Added python typing

* formatting

* fix: workflow params, localized try/catch, types, tidying

* refactor: activities to fit blog, workers to single aync function

* feat: sync return result

* refactor: tidy

* refactor: only generate summary at wf end

* fix: query named

* docs: add bedrock link to readme

* refactor: rename bedrock sample directories

* build: use poetry to run bedrock samples

* refactor: rename dirs to allow poe to run without modification

* refactor: shared activity, shared boto3 client, separate workflow ids

* refactor: remove unneeded kwargs

* refactor: minor tidyups

* fix: linting

* fix: start workflow with empty params

* refactor: only summarize once

* refactor: linting

* refactor

* log warning if messages received while chat is closed

* chore: linting

* Update bedrock/basic/README.md

Co-authored-by: Chad Retz <[email protected]>

* Update bedrock/basic/send_message.py

Co-authored-by: Chad Retz <[email protected]>

* Update bedrock/entity/workflows.py

Co-authored-by: Chad Retz <[email protected]>

* chad changes

* prerequisites in README

* refactor: mionr changes

---------

Co-authored-by: Steve Androulakis <[email protected]>
Co-authored-by: Chad Retz <[email protected]>

Author: 3 people

README.md

@@ -52,6 +52,7 @@ Some examples require extra dependencies. See each sample's directory for specif
   * [hello_signal](hello/hello_signal.py) - Send signals to a workflow.
 <!-- Keep this list in alphabetical order -->
 * [activity_worker](activity_worker) - Use Python activities from a workflow in another language.
+* [bedrock](bedrock) - Orchestrate a chatbot with Amazon Bedrock.
 * [cloud_export_to_parquet](cloud_export_to_parquet) - Set up schedule workflow to process exported files on an hourly basis
 * [context_propagation](context_propagation) - Context propagation through workflows/activities via interceptor.
 * [custom_converter](custom_converter) - Use a custom payload converter to handle custom types.

bedrock/README.md

@@ -0,0 +1,25 @@
+# AI Chatbot example using Amazon Bedrock
+
+Demonstrates how Temporal and Amazon Bedrock can be used to quickly build bulletproof AI applications.
+
+## Samples
+
+* [basic](basic) - A basic Bedrock workflow to process a single prompt.
+* [signals_and_queries](signals_and_queries) - Extension to the basic workflow to allow multiple prompts through signals & queries.
+* [entity](entity) - Full multi-Turn chat using an entity workflow..
+
+## Pre-requisites
+
+1. An AWS account with Bedrock enabled.
+2. A machine that has access to Bedrock.
+3. A local Temporal server running on the same machine. See [Temporal's dev server docs](https://docs.temporal.io/cli#start-dev-server) for more information.
+
+These examples use Amazon's Python SDK (Boto3). To configure Boto3 to use your AWS credentials, follow the instructions in [the Boto3 documentation](https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html).
+
+## Running the samples
+
+For these sample, the optional `bedrock` dependency group must be included. To include, run:
+
+    poetry install --with bedrock
+
+There are 3 Bedrock samples, see the README.md in each sub-directory for instructions on running each.

bedrock/__init__.py

@@ -0,0 +1,10 @@
+# Basic Amazon Bedrock workflow
+
+A basic Bedrock workflow. Starts a workflow with a prompt, generates a response and ends the workflow.
+
+To run, first see `samples-python` [README.md](../../README.md), and `bedrock` [README.md](../README.md) for prerequisites specific to this sample. Once set up, run the following from this directory:
+
+1. Run the worker: `poetry run python run_worker.py`
+2. In another terminal run the client with a prompt:
+
+    e.g. `poetry run python send_message.py 'What animals are marsupials?'`

bedrock/basic/README.md

@@ -0,0 +1,35 @@
+import asyncio
+import concurrent.futures
+import logging
+
+from temporalio.client import Client
+from temporalio.worker import Worker
+from workflows import BasicBedrockWorkflow
+
+from bedrock.shared.activities import BedrockActivities
+
+
+async def main():
+    # Create client connected to server at the given address
+    client = await Client.connect("localhost:7233")
+    activities = BedrockActivities()
+
+    # Run the worker
+    with concurrent.futures.ThreadPoolExecutor(max_workers=100) as activity_executor:
+        worker = Worker(
+            client,
+            task_queue="bedrock-task-queue",
+            workflows=[BasicBedrockWorkflow],
+            activities=[activities.prompt_bedrock],
+            activity_executor=activity_executor,
+        )
+        await worker.run()
+
+
+if __name__ == "__main__":
+    print("Starting worker")
+    print("Then run 'python send_message.py \"<prompt>\"'")
+
+    logging.basicConfig(level=logging.INFO)
+
+    asyncio.run(main())

bedrock/basic/__init__.py

@@ -0,0 +1,29 @@
+import asyncio
+import sys
+
+from temporalio.client import Client
+from workflows import BasicBedrockWorkflow
+
+
+async def main(prompt: str) -> str:
+    # Create client connected to server at the given address
+    client = await Client.connect("localhost:7233")
+
+    # Start the workflow
+    workflow_id = "basic-bedrock-workflow"
+    handle = await client.start_workflow(
+        BasicBedrockWorkflow.run,
+        prompt,  # Initial prompt
+        id=workflow_id,
+        task_queue="bedrock-task-queue",
+    )
+    return await handle.result()
+
+
+if __name__ == "__main__":
+    if len(sys.argv) != 2:
+        print("Usage: python send_message.py '<prompt>'")
+        print("Example: python send_message.py 'What animals are marsupials?'")
+    else:
+        result = asyncio.run(main(sys.argv[1]))
+        print(result)

bedrock/basic/run_worker.py

@@ -0,0 +1,24 @@
+from datetime import timedelta
+
+from temporalio import workflow
+
+with workflow.unsafe.imports_passed_through():
+    from bedrock.shared.activities import BedrockActivities
+
+
+@workflow.defn
+class BasicBedrockWorkflow:
+    @workflow.run
+    async def run(self, prompt: str) -> str:
+
+        workflow.logger.info("Prompt: %s" % prompt)
+
+        response = await workflow.execute_activity_method(
+            BedrockActivities.prompt_bedrock,
+            prompt,
+            schedule_to_close_timeout=timedelta(seconds=20),
+        )
+
+        workflow.logger.info("Response: %s" % response)
+
+        return response

bedrock/basic/send_message.py

@@ -0,0 +1,19 @@
+# Multi-turn chat with Amazon Bedrock Entity Workflow
+
+Multi-Turn Chat using an Entity Workflow. The workflow runs forever unless explicitly ended. The workflow continues as new after a configurable number of chat turns to keep the prompt size small and the Temporal event history small. Each continued-as-new workflow receives a summary of the conversation history so far for context.
+
+To run, first see `samples-python` [README.md](../../README.md), and `bedrock` [README.md](../README.md) for prerequisites specific to this sample. Once set up, run the following from this directory:
+
+1. Run the worker: `poetry run python run_worker.py`
+2. In another terminal run the client with a prompt.
+
+    Example: `poetry run python send_message.py 'What animals are marsupials?'`
+
+3. View the worker's output for the response.
+4. Give followup prompts by signaling the workflow.
+
+    Example: `poetry run python send_message.py 'Do they lay eggs?'`
+5. Get the conversation history summary by querying the workflow.
+    
+    Example: `poetry run python get_history.py`
+6. To end the chat session, run `poetry run python end_chat.py`