app.py

import os
from dotenv import load_dotenv
import gradio as gr
import chromadb
from chromadb.utils import embedding_functions
from langchain_openai import ChatOpenAI
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from pymongo import MongoClient
from pymongo.errors import ConnectionFailure
from urllib.parse import quote_plus

# Load environment variables
load_dotenv()
openai_api_key = os.getenv("OPENAI_API_KEY")
REPO_ID = "poemsforaphrodite/rag"  # Replace with your actual space ID
CHROMA_PATH = "./chroma_db"

# MongoDB connection details
MONGO_URI = os.getenv("MONGO_URI")

# Initialize MongoDB client
try:
    mongo_client = MongoClient(MONGO_URI)
    db = mongo_client.get_database("chatbot_db")
    chat_logs_collection = db.get_collection("chat_logs")
except ConnectionFailure as e:
    raise

def clean_text_with_gpt(text: str) -> str:
    model = ChatOpenAI(model="gpt-4o-mini", openai_api_key=openai_api_key)
    template = """Clean the following text, only show the main text
    
    {text}
    
    Cleaned text:"""
    prompt = ChatPromptTemplate.from_template(template)
    chain = prompt | model | StrOutputParser()
    return chain.invoke({"text": text})

def log_chat(query: str, response: str):
    chat_entry = {
        "messages": [
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": query},
            {"role": "assistant", "content": response}
        ]
    }
    try:
        chat_logs_collection.insert_one(chat_entry)
    except Exception:
        pass

def query_all_collections(query: str) -> tuple[str, str]:
    try:
        # Initialize Chroma client
        chroma_client = chromadb.PersistentClient(path=CHROMA_PATH)
        
        # Create an OpenAI embedding function
        openai_ef = embedding_functions.OpenAIEmbeddingFunction(
            api_key=openai_api_key,
            model_name="text-embedding-ada-002"
        )
        
        # Get all collection names
        collection_names = [col.name for col in chroma_client.list_collections()]
        
        all_results = []
        for collection_name in collection_names:
            try:
                collection = chroma_client.get_collection(name=collection_name, embedding_function=openai_ef)
                results = collection.query(
                    query_texts=[query],
                    n_results=1,  # Adjust as needed
                    include=['documents', 'metadatas', 'distances']
                )
                
                for doc, meta in zip(results['documents'][0], results['metadatas'][0]):
                    book_name = meta.get('file_name') or meta.get('source') or meta.get('book_name') or 'Unknown'
                    all_results.append((doc, book_name, collection_name))
            except Exception:
                continue
        
        if not all_results:
            return "No relevant documents found.", ""
        
        # Sort results by relevance (assuming the query returns most relevant first)
        all_results.sort(key=lambda x: x[0])
        
        # Combine results from all collections and clean the text using GPT-4
        combined_docs = []
        for i, (doc, book_name, col_name) in enumerate(all_results):
            cleaned_text = clean_text_with_gpt(doc)
            formatted_doc = f"""
Document {i+1}:
Collection: {col_name}
Book: {os.path.basename(book_name)}
Content: {cleaned_text}
---"""
            combined_docs.append(formatted_doc)
        
        context = "\n".join([doc for doc, _, _ in all_results])
        
        # Use the combined results for the response
        model = ChatOpenAI(model="gpt-4o-mini", openai_api_key=openai_api_key)
        template = """Answer the question based only on the following context:
        {context}
        Question: {question}
        
        After providing your answer, please add the following question:
        "Based on your clinical judgment and the patient's complete history and current presentation, do you agree with this recommendation, or are there additional considerations or adjustments needed?"
        """
        prompt = ChatPromptTemplate.from_template(template)
        chain = prompt | model | StrOutputParser()
        response = chain.invoke({"context": context, "question": query})
        
        response_text = f"Response: {response}"
        
        # Log the chat interaction
        log_chat(query, response)
        
        return response_text, "\n".join(combined_docs)
    except Exception as e:
        return f"An error occurred: {str(e)}", ""

# Gradio interface
iface = gr.Interface(
    fn=query_all_collections,
    inputs=[
        gr.Textbox(lines=1, placeholder="Enter your query here")
    ],
    outputs=[
        gr.Textbox(lines=10, label="Answer"),
        gr.Textbox(lines=10, label="Relevant Document Chunks")
    ],
    title="Multi-Collection Document Retrieval and QA Chatbot",
    description="Ask questions based on the content across all collections in the document database.",
)

if __name__ == "__main__":
    iface.launch(share=True)