app.py

import os
import gradio as gr
import qdrant_client
from llama_index.core import Settings, VectorStoreIndex, StorageContext
from llama_index.vector_stores.qdrant import QdrantVectorStore
from llama_index.embeddings.fastembed import FastEmbedEmbedding
from llama_index.llms.gemini import Gemini
from llama_index.core.memory import ChatMemoryBuffer
from llama_index.readers.web import FireCrawlWebReader
import dotenv
import time

dotenv.load_dotenv()

# Global variables
index = None
chat_engine = None
collection_name = ""

def embed_setup():
    Settings.embed_model = FastEmbedEmbedding(model_name="BAAI/bge-small-en-v1.5")
    Settings.llm = Gemini(temperature=0.1, model_name="models/gemini-pro")

def qdrant_setup():
    client = qdrant_client.QdrantClient(
        os.getenv("QDRANT_URL"),
        api_key=os.getenv("QDRANT_API_KEY"),
    )
    return client

def ingest_documents(url):
    firecrawl_reader = FireCrawlWebReader(
        api_key=os.getenv("FIRECRAWL_API_KEY"),
        mode="scrape",
    )
    documents = firecrawl_reader.load_data(url=url)
    return documents

def setup_query_engine(url, coll_name):
    global index, chat_engine, collection_name
    collection_name = coll_name
    
    embed_setup()
    client = qdrant_setup()
    vector_store = QdrantVectorStore(client=client, collection_name=collection_name)
    storage_context = StorageContext.from_defaults(vector_store=vector_store)
    
    if url:
        documents = ingest_documents(url)
        index = VectorStoreIndex.from_documents(documents, vector_store=vector_store, storage_context=storage_context)
    else:
        index = VectorStoreIndex.from_vector_store(vector_store=vector_store, storage_context=storage_context)
    
    memory = ChatMemoryBuffer.from_defaults(token_limit=4000)
    chat_engine = index.as_chat_engine(
        chat_mode="context",
        memory=memory,
        system_prompt=(
            """You are an AI assistant for developers, specializing in technical documentation. Your task is to provide accurate, concise, and helpful responses based on the given documentation context.
            Context information is below:
            {context_str}
            Always answer based on the information in the context and general knowledge and be precise
            Given this context, please respond to the following user query:
            {query_str}
            Your response should:
            Directly address the query using information from the context
            Include relevant code examples or direct quotes if applicable
            Mention specific sections or pages of the documentation
            Highlight any best practices or potential pitfalls related to the query
            After your response, suggest 3 follow-up questions based on the context that the user might find helpful for deeper understanding.
            ALWAYS SUGGEST FOLLOW UP QUESTIONS
            Your response:"""
        ),
    )
    return "Query engine setup completed successfully!"

def query_documentation(query):
    global chat_engine
    if not chat_engine:
        return "Please set up the query engine first."
    
    try:
        response = chat_engine.chat(query)
        return str(response.response)
    except Exception as e:
        error_message = f"An error occurred: {str(e)}"
        time.sleep(120)
        try:
            response = chat_engine.chat(query)
            return str(response.response)
        except Exception as e:
            return f"Retry failed. Error: {str(e)}"

# Gradio interface
with gr.Blocks() as app:
    gr.Markdown("# Talk to Software Documentation")
    
    with gr.Tab("Setup"):
        url_input = gr.Textbox(label="Enter URL to crawl and ingest documents (optional)")
        collection_input = gr.Textbox(label="Enter collection name for vector store (compulsory)")
        setup_button = gr.Button("Setup Query Engine")
        setup_output = gr.Textbox(label="Setup Output")
        
        setup_button.click(setup_query_engine, inputs=[url_input, collection_input], outputs=setup_output)
    
    with gr.Tab("Chat"):
        chatbot = gr.Chatbot()
        msg = gr.Textbox(label="Enter your query")
        clear = gr.Button("Clear")

        def user(user_message, history):
            return "", history + [[user_message, None]]

        def bot(history):
            user_message = history[-1][0]
            bot_message = query_documentation(user_message)
            history[-1][1] = bot_message
            return history

        msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
            bot, chatbot, chatbot
        )
        clear.click(lambda: None, None, chatbot, queue=False)

if __name__ == "__main__":
    app.launch()