app.py

# app.py
import os
from dotenv import load_dotenv
from pydantic import BaseModel
from qdrant_search import QdrantSearch
from langchain_groq import ChatGroq
from nomic_embeddings import EmbeddingsModel
import gradio as gr

load_dotenv()

import warnings
warnings.filterwarnings("ignore", category=FutureWarning)

os.environ["TOKENIZERS_PARALLELISM"] = "FALSE"

# Initialize global variables
collection_names = ["docs_v1_2", "docs_v2_2", "docs_v3_2"]
limit = 5
llm = ChatGroq(model="mixtral-8x7b-32768")
embeddings = EmbeddingsModel()
search = QdrantSearch(
    qdrant_url=os.environ["QDRANT_CLOUD_URL"],
    api_key=os.environ["QDRANT_API_KEY"],
    embeddings=embeddings
)

# Define the query processing function
def chat_with_langassist(query: str):
    if not query.strip():
        return "Query cannot be empty.", []
    
    # Retrieve relevant documents from Qdrant
    retrieved_docs = search.query_multiple_collections(query, collection_names, limit)
    
    # Prepare the context from retrieved documents
    context = "\n".join([doc['text'] for doc in retrieved_docs])
    
    # Construct the prompt with context and question
    prompt = (
        "You are LangAssist, a knowledgeable assistant for the LangChain Python Library. "
        "Given the following context from the documentation, provide a helpful answer to the user's question.\n\n"
        "Context:\n{context}\n\n"
        "Question: {question}\n\n"
        "Answer:"
    ).format(context=context, question=query)
    
    # Generate an answer using the language model
    try:
        answer = llm.invoke(prompt).content.strip()
    except Exception as e:
        return f"Error: {str(e)}", []
    
    # Prepare sources
    sources = [
        {
            "source": doc['source'],
            "text": doc['text']
        } for doc in retrieved_docs
    ]
    
    return answer, sources

# Define Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("<h1>LangAssist Chat</h1>")
    chatbot = gr.Chatbot()
    msg = gr.Textbox()
    clear = gr.Button("Clear")

    def respond(message, chat_history):
        answer, sources = chat_with_langassist(message)
        chat_history.append((message, answer))
        return chat_history, gr.update(value=''), sources

    msg.submit(respond, [msg, chatbot], [chatbot, msg])
    clear.click(lambda: None, None, chatbot)

# Run the Gradio app
if __name__ == "__main__":
    demo.launch()