2-notebooks/4-frameworks/2-autogen-multi-agent-rag.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "title-cell",
   "metadata": {},
   "source": [
    "# 🏋️‍♂️ Multi-Agent RAG for Fitness & Health with Azure AI Foundry 🥑\n",
    "\n",
    "Welcome to this fun, self-guided workshop where we'll build a multi-agent Retrieval-Augmented Generation (RAG) pipeline using AutoGen 0.4.7 with Azure AI Foundry. Our team of agents will collaborate to answer your fitness and health questions in an engaging way!"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "setup-markdown",
   "metadata": {},
   "source": [
    "## 1. Setup\n",
    "\n",
    "Let's import the necessary libraries and set up our model client using Azure AI Foundry. Make sure your environment variable `GITHUB_TOKEN` is set with your personal access token."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "setup-code",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import asyncio\n",
    "\n",
    "# Import AutoGen agents and required modules\n",
    "from autogen_agentchat.agents import AssistantAgent\n",
    "from autogen_agentchat.teams import RoundRobinGroupChat\n",
    "from autogen_agentchat.messages import TextMessage\n",
    "from autogen_core import CancellationToken\n",
    "\n",
    "# Import the Azure AI Foundry model client from AutoGen extensions\n",
    "from autogen_ext.models.azure import AzureAIChatCompletionClient\n",
    "from autogen_core.models import UserMessage\n",
    "from azure.core.credentials import AzureKeyCredential\n",
    "\n",
    "# Create the model client using Azure AI Foundry\n",
    "model_client = AzureAIChatCompletionClient(\n",
    "    model=os.environ[\"MODEL_DEPLOYMENT_NAME\"],\n",
    "    endpoint=\"https://models.inference.ai.azure.com\",\n",
    "    credential=AzureKeyCredential(os.environ[\"GITHUB_TOKEN\"]),\n",
    "    model_info={\n",
    "        \"json_output\": False,\n",
    "        \"function_calling\": False,\n",
    "        \"vision\": False,\n",
    "        \"family\": \"unknown\"\n",
    "    }\n",
    ")\n",
    "\n",
    "print(\"✅ Azure AI Foundry model client created successfully!\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "health-data-markdown",
   "metadata": {},
   "source": [
    "## 2. Create Sample Health Data and Retrieval Tool\n",
    "\n",
    "We'll define a small list of health tips and a simple retrieval function. This function simulates retrieving relevant health tips based on keywords in the user's query."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "health-data-code",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define sample health tips\n",
    "health_tips = [\n",
    "    {\"id\": \"tip1\", \"content\": \"Do a 10-minute HIIT workout to boost your metabolism.\", \"source\": \"Fitness Guru\"},\n",
    "    {\"id\": \"tip2\", \"content\": \"Take a brisk 15-minute walk to clear your mind and improve circulation.\", \"source\": \"Health Coach\"},\n",
    "    {\"id\": \"tip3\", \"content\": \"Stretch for 5 minutes every hour if you're sitting at a desk.\", \"source\": \"Wellness Expert\"},\n",
    "    {\"id\": \"tip4\", \"content\": \"Incorporate strength training twice a week for overall fitness.\", \"source\": \"Personal Trainer\"},\n",
    "    {\"id\": \"tip5\", \"content\": \"Drink water regularly to stay hydrated during workouts.\", \"source\": \"Nutritionist\"}\n",
    "]\n",
    "\n",
    "def retrieve_tips(query: str) -> str:\n",
    "    \"\"\"Return health tips whose content contains keywords from the query.\"\"\"\n",
    "    query_lower = query.lower()\n",
    "    relevant = []\n",
    "    for tip in health_tips:\n",
    "        # Check if any word in the query is in the tip content\n",
    "        if any(word in tip[\"content\"].lower() for word in query_lower.split()):\n",
    "            relevant.append(f\"Source: {tip['source']} => {tip['content']}\")\n",
    "    if not relevant:\n",
    "        # If no tips match, return all tips (for demo purposes)\n",
    "        relevant = [f\"Source: {tip['source']} => {tip['content']}\" for tip in health_tips]\n",
    "    return \"\\n\".join(relevant)\n",
    "\n",
    "print(\"✅ Sample health tips and retrieval tool created!\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "pipeline-markdown",
   "metadata": {},
   "source": [
    "## 3. Define Our Multi-Agent RAG Pipeline\n",
    "\n",
    "We will create two agents:\n",
    "\n",
    "1. **RetrieverAgent**: Uses the `retrieve_tips` tool to fetch relevant fitness and health tips based on the user's query.\n",
    "2. **ResponderAgent**: Crafts a fun, engaging response using the retrieved tips along with the user's query.\n",
    "\n",
    "These agents will be combined in a RoundRobinGroupChat to simulate collaborative problem-solving."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "pipeline-code",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the RetrieverAgent with the retrieve_tips tool\n",
    "retriever = AssistantAgent(\n",
    "    name=\"RetrieverAgent\",\n",
    "    system_message=(\n",
    "        \"You are a smart retrieval agent. Your task is to fetch relevant fitness and health tips based on the user's query. \"\n",
    "        \"Use the provided tool 'retrieve_tips' to get the information.\"\n",
    "    ),\n",
    "    model_client=model_client,\n",
    "    tools=[retrieve_tips],  # Register the retrieval tool\n",
    "    reflect_on_tool_use=True\n",
    ")\n",
    "\n",
    "# Define the ResponderAgent which uses the retrieved tips to craft a response\n",
    "responder = AssistantAgent(\n",
    "    name=\"ResponderAgent\",\n",
    "    system_message=(\n",
    "        \"You are a friendly fitness coach. Use the health tips provided to craft a fun and engaging response to the user's question.\"\n",
    "    ),\n",
    "    model_client=model_client\n",
    ")\n",
    "\n",
    "# Combine the two agents in a RoundRobinGroupChat\n",
    "group_chat = RoundRobinGroupChat(\n",
    "    agents=[retriever, responder],\n",
    "    termination_condition=None,  # For simplicity, we run a fixed number of turns\n",
    "    max_turns=4  # Limit the conversation to 4 turns\n",
    ")\n",
    "\n",
    "print(\"✅ Multi-agent RAG pipeline defined!\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "try-query-markdown",
   "metadata": {},
   "source": [
    "## 4. Try a Query\n",
    "\n",
    "Let's test our multi-agent RAG system with a fun fitness query. For example:\n",
    "\n",
    "> **User Query:** _I'm very busy but want to stay fit. What quick exercises can I do?_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "try-query-code",
   "metadata": {},
   "outputs": [],
   "source": [
    "async def run_query():\n",
    "    user_query = \"I'm very busy but want to stay fit. What quick exercises can I do?\"\n",
    "\n",
    "    # Run the group chat to process the query\n",
    "    final_result = None\n",
    "    async for message in group_chat.run_stream(task=user_query):\n",
    "        # Print each message as it is produced\n",
    "        if hasattr(message, 'content'):\n",
    "            print(f\"{message.source}: {message.content}\")\n",
    "        final_result = message\n",
    "    \n",
    "    print(\"\\n✅ Final response:\", final_result.content if final_result else \"No response\")\n",
    "\n",
    "asyncio.run(run_query())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "conclusion-markdown",
   "metadata": {},
   "source": [
    "## 5. Conclusion\n",
    "\n",
    "In this notebook, we built a fun multi-agent Retrieval-Augmented Generation pipeline using AutoGen 0.4.7 with Azure AI Foundry and a fitness and health theme. We created a RetrieverAgent that fetches relevant health tips and a ResponderAgent that crafts an engaging answer to the user's query.\n",
    "\n",
    "Feel free to modify and expand this notebook to explore more advanced multi-agent collaborations. Happy coding and stay fit! 💪🥦"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.11.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}