app.py

import os
import streamlit as st
from crewai import Agent, Task, Crew
from crewai_tools import LlamaIndexTool
from langchain_groq import ChatGroq
from llama_index.core import SimpleDirectoryReader, VectorStoreIndex
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
import tempfile
import requests

# --- Streamlit UI Header ---
st.title("Document Q&A Assistant with CrewAI")
st.write("Upload a document, provide a link, or ask questions dynamically!")

# --- Key Configuration from Secrets ---
try:
    GROQ_API_KEY = st.secrets["GROQ_API_KEY"]
    TAVILY_API_KEY = st.secrets["TAVILY_API_KEY"]
except KeyError as e:
    st.error(f"Missing API key in secrets: {e}. Please add it to your environment.")
    st.stop()

# Check if all API keys are available
if not GROQ_API_KEY or not TAVILY_API_KEY:
    st.error("One or more required API keys are missing. Please check your configuration.")
    st.stop()

# Function to download PDF from URL
def download_pdf_from_url(url, save_path):
    response = requests.get(url)
    if response.status_code == 200:
        with open(save_path, 'wb') as f:
            f.write(response.content)
        return save_path
    else:
        st.error("Failed to download PDF from the provided URL.")
        return None

# --- User Inputs for File or Link ---
document_source = st.radio("Choose input method:", ("Upload a PDF", "Provide PDF URL"))

pdf_path = None
if document_source == "Upload a PDF":
    uploaded_file = st.file_uploader("Upload a PDF file", type=['pdf'])
    if uploaded_file:
        with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as temp_file:
            temp_file.write(uploaded_file.getvalue())
            pdf_path = temp_file.name
        st.success("File uploaded successfully!")
else:
    pdf_url = st.text_input("Enter PDF URL")
    if st.button("Download PDF") and pdf_url:
        with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as temp_file:
            saved_path = download_pdf_from_url(pdf_url, temp_file.name)
            if saved_path:
                pdf_path = saved_path

# --- LLM Configuration ---
llm = ChatGroq(groq_api_key=GROQ_API_KEY, model="groq/llama-3.3-70b-versatile")

# Function to create Query Engine
def create_query_engine(pdf_path, llm):
    reader = SimpleDirectoryReader(input_files=[pdf_path])
    docs = reader.load_data()
    embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")
    index = VectorStoreIndex.from_documents(docs, embed_model=embed_model)
    return index.as_query_engine(similarity_top_k=5)

# --- Streamlit Question Workflow ---
if pdf_path:
    st.success("PDF loaded successfully!")
    query_engine = create_query_engine(pdf_path, llm)
    query_tool = LlamaIndexTool.from_query_engine(
        query_engine,
        name="Document Query Tool",
        description="Tool to analyze and retrieve information from the uploaded document."
    )

    # Define Agents and Tasks
    researcher = Agent(
        role="Document Analyst",
        goal="Analyze documents and answer questions",
        backstory="Expert at retrieving insights from documents.",
        verbose=True,
        allow_delegation=False,
        tools=[query_tool],
        llm=llm,
    )

    task = Task(
        description="Answer user queries based on the uploaded document.",
        expected_output="Clear and concise answers to user questions.",
        agent=researcher,
    )

    crew = Crew(agents=[researcher], tasks=[task], verbose=True)

    st.subheader("Ask a Question")
    user_question = st.text_input("Enter your question")

    if st.button("Get Answer"):
        with st.spinner("Processing your request..."):
            result = crew.kickoff(inputs={"question": user_question})
            st.success("Here is the answer:")
            st.write(result)
else:
    st.warning("Please upload a PDF or provide a valid URL to continue.")

# --- Clean Up ---
if pdf_path and os.path.exists(pdf_path):
    os.remove(pdf_path)