app.py

import chainlit as cl
from langchain.agents import create_agent
from langchain.chat_models import init_chat_model
from langchain_core.runnables import Runnable, RunnableConfig
from langchain_core.vectorstores import InMemoryVectorStore
from langchain_community.document_loaders import DirectoryLoader, PyPDFLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
from huggingface_hub import snapshot_download
from langchain.tools import tool
from langchain_groq import ChatGroq
from typing import cast
import os

from dotenv import load_dotenv

load_dotenv()

from langchain_google_genai import GoogleGenerativeAIEmbeddings

embedding_model = GoogleGenerativeAIEmbeddings(model="models/gemini-embedding-001")
vector_store = InMemoryVectorStore(embedding_model)


def load_sys_prompt(file_path="system_message.txt"):
    with open(file_path, "r") as f:
        return f.read().strip()


system_prompt = load_sys_prompt()


@tool(response_format="content_and_artifact")
def retrieve_context(query: str) -> str:
    """Call this tool ONLY when the user asks specific questions about Jake's
    career, skills, experience, or resume.

    DO NOT call this tool for greetings, small talk, or general questions.
    """

    print(f"DEBUG: Tool called with query: {query}")  # See this in your terminal
    retrieved_docs = vector_store.similarity_search(query, k=2)

    if not retrieved_docs:
        print("DEBUG: No documents found in vector store!")
        return "No relevant documents found.", []

    print(f"{len(retrieved_docs)} document(s) found in vector store.")
    print(f"First document: {retrieved_docs[0].page_content[:50]}...")

    serialized = "\n\n".join(
        (f"Source: {doc.metadata}\nContent: {doc.page_content}")
        for doc in retrieved_docs
    )
    return serialized, retrieved_docs


@cl.on_chat_start
async def on_chat_start():

    try:
        # download rag data files from private HF dataset
        snapshot_download(
            repo_id="jakewatson91/sherlock-rag-docs",
            repo_type="dataset",
            local_dir="data/",
            allow_patterns="*.pdf",
            token=True,
        )
    except Exception as e:
        print(f"Error downloading data: {e}")

    loader = DirectoryLoader(
        "data/",
        glob="*.pdf",
        loader_cls=PyPDFLoader,
    )

    docs = loader.load()
    print(f"Loaded {len(docs)} documents")

    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)

    llm = init_chat_model(
        model="moonshotai/kimi-k2-instruct-0905",
        model_provider="groq",
        streaming=True,
        temperature=0,
    )

    runnable = create_agent(
        model=llm, tools=[retrieve_context], system_prompt=system_prompt
    )

    cl.user_session.set("runnable", runnable)
    cl.user_session.set("memory", [])


# Send a response back to the user
@cl.on_message
async def on_message(message: cl.Message):
    runnable = cast(Runnable, cl.user_session.get("runnable"))  # type: Runnable
    res = cl.Message(content="")
    cb = cl.LangchainCallbackHandler()

    memory = cl.user_session.get("memory")
    memory.append({"role": "user", "content": message.content})

    async for msg, metadata in runnable.astream(
        {"messages": memory},
        stream_mode="messages",
        config=RunnableConfig(callbacks=[cb], run_name="Sherlock Search"),
    ):
        if (
            msg.content
            and metadata.get("langgraph_node") == "model"
            and not getattr(msg, "tool_calls", None)
        ):
            print("METADATA: ", metadata)
            print("MESSAGE: ", msg)

            await res.stream_token(msg.content)

    memory.append({"role": "assistant", "content": res.content})
    cl.user_session.set("memory", memory)

    await res.send()


if __name__ == "__main__":
    snapshot_download(
        repo_id="jakewatson91/sherlock-rag-docs",
        repo_type="dataset",
        local_dir="data/",
        allow_patterns="*.pdf",
        token=os.environ.get("HF_TOKEN"),
    )