Update README.md

acp-code/README.md

@@ -73,7 +73,86 @@ A general summary from the first agent
 
 A fact-checked and updated version from the second agent
 
+
+## ACP-Code Repository Explanation
+
+This repository demonstrates the Agent Communication Protocol (ACP) - a standardized way for AI agents built with different frameworks to communicate and collaborate. Here's a breakdown:
+Overview
+The project shows how two agents using different frameworks (CrewAI and Smolagents) work together to:
+Generate a research summary on a topic
+Fact-check and enhance it using web search
+
+## Core Components
+1. CrewAI ACP Server (crew_acp_server.py)
+
+@server.agent()
+async def research_drafter(input: list[Message]) -> AsyncGenerator[RunYield, RunYieldResume]:
+Purpose: Creates initial research summaries Key elements:
+Uses Ollama's Qwen2.5:14b model (line 8-12)
+Defines a CrewAI Agent with role "Research summarizer" (lines 18-23)
+Creates a Task to write a brief summary (lines 25-29)
+Executes the crew and returns the result as an ACP Message (lines 31-33)
+Runs on port 8000 (line 36)
+How it works: Takes a topic as input, uses CrewAI's agent to generate a summary, and yields it back through ACP protocol.
+2. Smolagents ACP Server (smolagents_acp_server.py)
+
+@server.agent()
+async def research_verifier(input: list[Message]) -> AsyncGenerator[RunYield, RunYieldResume]:
+Purpose: Fact-checks and enhances summaries using web search Key elements:
+Uses the same Ollama model via LiteLLM (lines 11-16)
+Creates a CodeAgent with DuckDuckGo search tool (line 22)
+Takes the draft summary and searches for updated information (lines 24-25)
+Returns enhanced version (line 27)
+Runs on port 8001 (line 30)
+How it works: Receives a draft summary, uses DuckDuckGo to find current information, and returns a fact-checked, enhanced version.
+3. ACP Client (acp_client.py)
+
+async def run_workflow() -> None:
+Purpose: Orchestrates the multi-agent workflow Workflow:
+Connects to both servers (lines 5-6)
+drafter client → CrewAI server (port 8000)
+verifier client → Smolagents server (port 8001)
+Step 1 - Generate draft (lines 9-14):
+Sends topic to research_drafter agent
+Receives and prints draft summary
+Step 2 - Enhance draft (lines 16-21):
+Sends draft to research_verifier agent
+Agent searches web for latest info
+Receives and prints final enhanced summary
+
+## Architecture Diagram
+
+┌─────────────┐
+│   Client    │
+│ (acp_client)│
+└──────┬──────┘
+       │
+       ├──────────────┐
+       │              │
+       ▼              ▼
+┌──────────────┐  ┌──────────────┐
+│ CrewAI Agent │  │Smolagent Agent│
+│   (Port 8000)│  │   (Port 8001) │
+│              │  │               │
+│ Research     │  │ Research      │
+│ Drafter      │  │ Verifier      │
+│              │  │ + DuckDuckGo  │
+└──────┬───────┘  └──────┬────────┘
+       │                 │
+       └────────┬────────┘
+                ▼
+       Ollama (Qwen2.5:14b)
+
+## Key Concepts
+
+ACP SDK: Provides standardized communication (Messages, MessageParts) between agents regardless of framework
+Framework Agnostic: CrewAI and Smolagents can collaborate seamlessly
+Async Architecture: Uses Python async/await for efficient I/O operations
+Tool Integration: Smolagents agent has access to DuckDuckGo for real-time information
+
+
 ## 📬 Stay Updated with Our Newsletter!
+
 **Get a FREE Data Science eBook** 📖 with 150+ essential lessons in Data Science when you subscribe to our newsletter! Stay in the loop with the latest tutorials, insights, and exclusive resources. [Subscribe now!](https://join.dailydoseofds.com)
 
 [![Daily Dose of Data Science Newsletter](https://github.com/patchy631/ai-engineering/blob/main/resources/join_ddods.png)](https://join.dailydoseofds.com)