Practical Workshop: Create, Test, and Deploy an Open-Source AI Agent Using LangChain

Objective

By the end of this workshop, participants will have created, tested, and deployed an open-source AI agent using LangChain. This workshop includes hands-on activities, from setup to deployment, designed for beginners and intermediate developers.

Agenda

1. Introduction to LangChain (15 minutes)

• Overview of LangChain and its capabilities.
• Use cases of AI agents.
• Workshop goals.

2. Environment Setup (15 minutes)

• Install Python and necessary tools.
• Install LangChain and dependencies.
• Verify setup.

3. Creating an AI Agent with LangChain (45 minutes)

• Build an agent capable of answering questions using an LLM and a document database.

4. Testing the AI Agent (30 minutes)

• Test the agent locally with sample inputs.
• Debugging and improving the agent’s performance.

5. Deploying the AI Agent (45 minutes)

• Deploy as an API using FastAPI.
• Test the API with a REST client (e.g., Postman).

6. Conclusion and Q&A (15 minutes)

• Review key takeaways.
• Discuss potential use cases and enhancements.
• Open floor for questions.

Detailed Steps

1. Environment Setup

Requirements:

• Python 3.8 or later
• VS Code or any text editor
• Postman or cURL for API testing

Commands:

# Install LangChain and dependencies
pip install langchain openai fastapi uvicorn

# Optional: Install libraries for document retrieval
pip install faiss-cpu unstructured

2. Create an AI Agent

Code Snippet: Basic LangChain Agent

from langchain.agents import initialize_agent, Tool
from langchain.chat_models import ChatOpenAI
from langchain.prompts import PromptTemplate

# Define tools
def search_tool(query: str) -> str:
    # Example tool: simple search function
    return f"Simulated search result for: {query}"

tools = [
    Tool(
        name="Search",
        func=search_tool,
        description="Use this tool to search for information."
    )
]

# Initialize LLM and agent
llm = ChatOpenAI(temperature=0, model="gpt-4")
prompt = PromptTemplate(template="Answer based on the tools: {query}", input_variables=["query"])

agent = initialize_agent(tools, llm, agent_type="zero-shot-react-description")

# Test agent
response = agent.run("Who is the president of the Philippines?")
print(response)

What This Does:

• Creates a zero-shot agent.
• Adds a custom search tool.
• Interacts with the agent using a natural language query.

3. Test the AI Agent

• Use diverse prompts to test the agent’s responses.
• Add more tools as needed, such as document retrieval or APIs.

4. Deploy the AI Agent

Code Snippet: API Deployment with FastAPI

from fastapi import FastAPI
from pydantic import BaseModel

app = FastAPI()

class Query(BaseModel):
    query: str

@app.post("/ask")
async def ask_agent(query: Query):
    response = agent.run(query.query)
    return {"response": response}

# Run with Uvicorn
# uvicorn filename:app --reload

Steps:

1. Save the file as app.py.
2. Run the server: uvicorn app:app --reload
3. Test with Postman or cURL: curl -X POST "http://127.0.0.1:8000/ask" -H "Content-Type: application/json" -d '{"query": "What is LangChain?"}'

5. Enhance the Agent

• Add a document database for knowledge retrieval (e.g., FAISS or Chroma).
• Integrate additional APIs for multi-functionality.

Materials

• LangChain Documentation
• Preloaded datasets for testing.
• Reference code snippets and slides (provided during the workshop).

Expected Outcome

Participants will:

1. Understand LangChain’s structure.
2. Build a working AI agent.
3. Deploy the agent as a REST API.
4. Gain insight into practical applications and potential enhancements.

Would you like additional resources, slides, or marketing materials for this workshop?