app.py

# ==============================================================================
# 1. SETUP AND IMPORTS
# ==============================================================================
import os
from fastapi import FastAPI
from pydantic import BaseModel
from typing import Optional

# Set Google API Key securely from an environment variable
google_api_key = os.getenv("GOOGLE_API_KEY")
if not google_api_key:
    raise ValueError("Google API key not found. Please set the GOOGLE_API_KEY environment variable.")

# All your other imports...
import bs4
from langchain import hub
from langchain_community.document_loaders import WebBaseLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_google_genai import ChatGoogleGenerativeAI, GoogleGenerativeAIEmbeddings
from langchain_core.vectorstores import InMemoryVectorStore
from langgraph.graph import MessagesState, StateGraph, END
from langgraph.prebuilt import ToolNode, tools_condition
from langchain_core.messages import HumanMessage
from langchain_core.tools import tool
from langgraph.checkpoint.memory import MemorySaver

# ==============================================================================
# 2. CORE LOGIC
# ==============================================================================
llm = ChatGoogleGenerativeAI(model="gemini-2.5-flash", google_api_key=google_api_key)
embeddings = GoogleGenerativeAIEmbeddings(model="models/embedding-001", google_api_key=google_api_key)
vector_store = InMemoryVectorStore(embeddings)

# Load Web Data and Split
web_url = "https://lilianweng.github.io/posts/2023-06-23-agent/"
loader = WebBaseLoader(
    web_paths=(web_url,),
    bs_kwargs=dict(parse_only=bs4.SoupStrainer(class_=("post-content", "post-title", "post-header")))
)
docs = loader.load()
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200, add_start_index=True)
all_splits = text_splitter.split_documents(docs)
vector_store.add_documents(all_splits)

# Tool Definition
@tool
def retrieve(query: str):
    """retrieve information related to a query."""
    retrieved_docs = vector_store.similarity_search(query, k=2)
    return [doc.page_content for doc in retrieved_docs]

# Graph Node Functions
def query_or_respond(state: MessagesState):
    llm_with_tools = llm.bind_tools([retrieve])
    response = llm_with_tools.invoke(state["messages"])
    return {"messages": [response]}

tools = ToolNode([retrieve])

def generate(state: MessagesState):
    response = llm.invoke(state["messages"])
    return {"messages": [response]}

# Compile the LangGraph StateGraph
graph_builder = StateGraph(MessagesState)
graph_builder.add_node("query_or_respond", query_or_respond)
graph_builder.add_node("tools", tools)
graph_builder.add_node("generate", generate)
graph_builder.set_entry_point("query_or_respond")
graph_builder.add_conditional_edges(
    "query_or_respond",
    tools_condition,
    {"tools": "tools", END: END}
)
graph_builder.add_edge("tools", "generate")
graph_builder.add_edge("generate", END)

memory = MemorySaver()
graph = graph_builder.compile(checkpointer=memory)

# ==============================================================================
# 3. API SERVER (Replaces your if __name__ == "__main__": block)
# ==============================================================================
app = FastAPI(
    title="LangGraph RAG Agent Server",
    description="An API server for a RAG agent built with LangGraph.",
)

# Define the input model for the API
class UserRequest(BaseModel):
    message: str
    thread_id: Optional[str] = "default_thread" # Use a default thread_id if none is provided

# Define the API endpoint
@app.post("/invoke")
async def invoke_agent(request: UserRequest):
    # Set up the configuration for memory
    config = {"configurable": {"thread_id": request.thread_id}}
    
    # Define the input for the graph
    inputs = {"messages": [HumanMessage(content=request.message)]}

    # Invoke the graph to get the final result
    response = graph.invoke(inputs, config=config)
    
    # Return the AI's final message
    final_message = response["messages"][-1]
    return {"response": final_message.content}

# This part is for local testing, can be removed if using a production server
if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)