src/streamlit_app.py

# 🫠 Clean and Final Streamlit RAG App (Three-Agent Architecture)

# --- Environment Setup (Safe for Hugging Face) ---
import os
os.environ["STREAMLIT_HOME"] = "/tmp"
os.environ["XDG_CONFIG_HOME"] = "/tmp"
os.environ["XDG_DATA_HOME"] = "/tmp"
os.environ["HOME"] = "/tmp"

import asyncio
try:
    asyncio.get_running_loop()
except RuntimeError:
    asyncio.set_event_loop(asyncio.new_event_loop())

# --- Imports ---
import streamlit as st
import pandas as pd
import json
import io
from dotenv import load_dotenv, find_dotenv
from langchain_openai import OpenAIEmbeddings, OpenAI
from langchain_community.vectorstores import FAISS
from langchain_experimental.agents.agent_toolkits import create_pandas_dataframe_agent
from langchain.text_splitter import CharacterTextSplitter
from langchain.chains import RetrievalQA
import PyPDF2
from docx import Document

# --- Load API Key Securely ---
_ = load_dotenv(find_dotenv())
try:
    API_KEY = st.secrets["OPENAI_API_KEY"]
except st.runtime.secrets.StreamlitSecretNotFoundError:
    API_KEY = os.getenv("OPENAI_API_KEY")

embeddings_model = OpenAIEmbeddings(openai_api_key=API_KEY)

# --- Streamlit Page Setup ---
st.set_page_config(page_title="RAG File Chat", layout="centered")
st.title("🧠 Chat with Your Uploaded File")

# --- Agent 1: File Ingestion and Indexing ---
def agent_alpha_file_uploader(file_content, file_type):
    st.info("📂 Agent 1: Loading and indexing your file...")
    if file_type == "csv":
        df = pd.read_csv(io.StringIO(file_content.decode("utf-8")))
        llm = OpenAI(openai_api_key=API_KEY)
        return create_pandas_dataframe_agent(llm, df, verbose=False), None
    elif file_type == "xlsx":
        df = pd.read_excel(file_content)
        llm = OpenAI(openai_api_key=API_KEY)
        return create_pandas_dataframe_agent(llm, df, verbose=False), None
    elif file_type == "json":
        df = pd.DataFrame(json.loads(file_content.decode("utf-8")))
        llm = OpenAI(openai_api_key=API_KEY)
        return create_pandas_dataframe_agent(llm, df, verbose=False), None
    elif file_type in ["pdf", "docx"]:
        text = extract_text_from_file(file_content, file_type)
        chunks = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0).split_text(text)
        vectorstore = FAISS.from_texts(chunks, embeddings_model)
        return None, vectorstore
    else:
        st.error("❌ Unsupported file type.")
        return None, None

# --- Agent 2: Query Resolution ---
def agent_beta_query_processor(query, file_type, df_agent=None, vectorstore=None):
    st.info("🧠 Agent 2: Processing your question...")
    if file_type in ["pdf", "docx"]:
        qa_chain = RetrievalQA.from_chain_type(
            llm=OpenAI(openai_api_key=API_KEY),
            chain_type="stuff",
            retriever=vectorstore.as_retriever(search_kwargs={"k": 5}),
        )
        result = qa_chain({"query": query})
        return result["result"]
    else:
        return df_agent.run(query)

# --- Agent 3: Response Enhancement ---
def agent_gamma_response_enhancer(response):
    st.info("🔍 Agent 3: Reviewing and enhancing the response...")
    enhancement_prompt = f"Improve the clarity and format of the following response:\n{response}"
    llm = OpenAI(openai_api_key=API_KEY)
    return llm.invoke(enhancement_prompt)

# --- Helper Function for Text Extraction ---
def extract_text_from_file(file_content, file_type):
    if file_type == "pdf":
        reader = PyPDF2.PdfReader(io.BytesIO(file_content))
        return "\n".join([page.extract_text() for page in reader.pages if page.extract_text()])
    elif file_type == "docx":
        doc = Document(io.BytesIO(file_content))
        return "\n".join([p.text for p in doc.paragraphs if p.text.strip()])
    return ""

# --- Session State ---
if "uploaded_file" not in st.session_state:
    st.session_state.uploaded_file = None
if "file_uploaded" not in st.session_state:
    st.session_state.file_uploaded = False
if "vectorstore" not in st.session_state:
    st.session_state.vectorstore = None
if "agent" not in st.session_state:
    st.session_state.agent = None
if "file_type" not in st.session_state:
    st.session_state.file_type = None

# --- File Upload UI ---
MAX_SIZE_MB = 50
uploaded = st.file_uploader("📁 Browse and select a file", type=["csv", "xlsx", "json", "pdf", "docx"])
if uploaded:
    st.write("🧾 Debug Upload Log:")
    st.code(f"Filename: {uploaded.name} | Type: {uploaded.type} | Size: {uploaded.size} bytes")
    st.write("📥 Upload Raw Details (client-side and browser environment check):")
    st.json({
        "user_agent": st.experimental_get_query_params().get('user-agent', ['N/A'])[0],
        "streamlit_version": st.__version__,
        "name": uploaded.name,
        "type": uploaded.type,
        "size_bytes": uploaded.size,
        "is_excel": uploaded.name.endswith(('.xls', '.xlsx')),
        "is_binary": isinstance(uploaded, io.BytesIO),
    })
    st.write("🧾 Debug Upload Log:")
    st.code(f"Filename: {uploaded.name} | Type: {uploaded.type} | Size: {uploaded.size} bytes")
    st.write("📝 File metadata:")
    st.json({"name": uploaded.name, "size_kb": uploaded.size / 1024, "type": uploaded.type})

    if uploaded.size > MAX_SIZE_MB * 1024 * 1024:
        st.error(f"❌ File too large. Maximum allowed size is {MAX_SIZE_MB}MB.")
        st.stop()
    st.session_state.uploaded_file = uploaded
    st.info(f"✅ File selected: `{uploaded.name}` ({uploaded.size / 1024:.1f} KB)")

if st.session_state.uploaded_file and st.button("📤 Upload File"):
    try:
        st.write("📡 Reading uploaded file content...")
        content = st.session_state.uploaded_file.read()
        st.write("🧪 File read complete. Size:", len(content), "bytes")
        st.code(str(content[:200]) + "..." if isinstance(content, bytes) else content[:500])
        ftype = st.session_state.uploaded_file.name.split(".")[-1].lower()
        with st.spinner("🔄 Agent Alpha (Uploader): Processing and indexing file..."):
            agent, vectorstore = agent_alpha_file_uploader(content, ftype)
            if agent or vectorstore:
                st.session_state.agent = agent
                st.session_state.vectorstore = vectorstore
                st.session_state.file_uploaded = True
                st.session_state.file_type = ftype
                st.success("✅ File processed successfully.")
            else:
                st.error("⚠️ Failed to process file.")
    except Exception as e:
        st.error("❌ Upload failed. Try a smaller file or check connection.")
        st.exception(e)

# --- Query UI ---
if st.session_state.file_uploaded:
    output_format = st.selectbox("📋 Select Output Format", ["Plain Text", "Markdown", "Tabular View"])
    query = st.text_area("🔍 Ask a question about your uploaded file")

    if st.button("Submit Query"):
        if not query.strip():
            st.warning("⚠️ Please enter a valid question.")
        else:
            with st.spinner("💡 Agent Beta (Processor): Handling your query..."):
                raw_response = agent_beta_query_processor(
                    query,
                    st.session_state.file_type,
                    df_agent=st.session_state.agent,
                    vectorstore=st.session_state.vectorstore,
                )

            with st.spinner("✨ Agent Gamma (Reviewer): Enhancing the response..."):
                enhanced_response = agent_gamma_response_enhancer(raw_response)

            st.subheader("📌 Final Answer")
            if output_format == "Plain Text":
                st.text(enhanced_response)
            elif output_format == "Markdown":
                st.markdown(enhanced_response)
            elif output_format == "Tabular View":
                rows = [line.split("\t") for line in enhanced_response.split("\n") if "\t" in line]
                if not rows or len(rows[0]) == 1:
                    rows = [line.split(",") for line in enhanced_response.split("\n") if "," in line]
                try:
                    df = pd.DataFrame(rows[1:], columns=rows[0])
                    st.dataframe(df)
                except Exception:
                    st.warning("⚠️ Could not render table. Showing raw text.")
                    st.text(enhanced_response)