app2.py

import os
import streamlit as st
import faiss
import pickle
import asyncio
import aiohttp
from huggingface_hub import hf_hub_download
from sentence_transformers import SentenceTransformer
import feedparser
import datetime
import numpy as np

# -------------------
# Streamlit config
# -------------------
st.set_page_config(page_title="📊 Agentic Financial RAG", layout="wide")

# -------------------
# Load FAISS index + metadata
# -------------------
@st.cache_resource
def load_finance_index():
    faiss_path = hf_hub_download(
        repo_id="krishnasimha/finance_data",
        filename="financial_phrasebank_index.faiss",
        repo_type="dataset"
    )
    pkl_path = hf_hub_download(
        repo_id="krishnasimha/finance_data",
        filename="financial_phrasebank_metadata.pkl",
        repo_type="dataset"
    )

    index = faiss.read_index(faiss_path)
    with open(pkl_path, "rb") as f:
        metadata = pickle.load(f)

    embed_model = SentenceTransformer("all-MiniLM-L6-v2")
    return index, metadata, embed_model

index, metadata, embed_model = load_finance_index()

# -------------------
# FAISS retrieval
# -------------------
def retrieve_docs(query, k=3):
    query_emb = embed_model.encode([query], convert_to_numpy=True)
    D, I = index.search(query_emb, k)
    retrieved_texts = [metadata["texts"][i] for i in I[0]]
    retrieved_sources = [metadata["sources"][i] for i in I[0]]
    return retrieved_texts, retrieved_sources

# -------------------
# Async Together API call
# -------------------
async def async_together_chat(messages):
    url = "https://api.together.xyz/v1/chat/completions"
    headers = {
        "Authorization": f"Bearer {os.environ['TOGETHER_API_KEY']}",
        "Content-Type": "application/json",
    }
    payload = {
        "model": "deepseek-ai/DeepSeek-V3",
        "messages": messages,
    }
    async with aiohttp.ClientSession() as session:
        async with session.post(url, headers=headers, json=payload) as resp:
            result = await resp.json()
            return result["choices"][0]["message"]["content"]

# -------------------
# RSS Financial News
# -------------------
RSS_URL = "https://news.google.com/rss/search?q=finance+stock+market&hl=en-IN&gl=IN&ceid=IN:en"

async def fetch_rss_url(url):
    async with aiohttp.ClientSession() as session:
        async with session.get(url) as resp:
            return await resp.text()

def get_financial_news():
    raw_xml = asyncio.run(fetch_rss_url(RSS_URL))
    feed = feedparser.parse(raw_xml)
    articles = []
    for entry in feed.entries[:5]:
        articles.append({
            "title": entry.title,
            "link": entry.link,
            "published": entry.published
        })
    return articles

# -------------------
# Session State: Chats
# -------------------
if "chats" not in st.session_state:
    st.session_state.chats = {}
if "current_chat" not in st.session_state:
    st.session_state.current_chat = "New Chat 1"
    st.session_state.chats["New Chat 1"] = [
        {"role": "system", "content": "You are a helpful financial assistant."}
    ]

# -------------------
# Sidebar: Chat Manager
# -------------------
st.sidebar.header("Chat Manager")
if st.sidebar.button("➕ New Chat"):
    chat_count = len(st.session_state.chats) + 1
    new_chat_name = f"New Chat {chat_count}"
    st.session_state.chats[new_chat_name] = [
        {"role": "system", "content": "You are a helpful financial assistant."}
    ]
    st.session_state.current_chat = new_chat_name

chat_list = list(st.session_state.chats.keys())
selected_chat = st.sidebar.selectbox("Your chats:", chat_list, index=chat_list.index(st.session_state.current_chat))
st.session_state.current_chat = selected_chat

new_name = st.sidebar.text_input("Rename Chat:", st.session_state.current_chat)
if new_name and new_name != st.session_state.current_chat:
    if new_name not in st.session_state.chats:
        st.session_state.chats[new_name] = st.session_state.chats.pop(st.session_state.current_chat)
        st.session_state.current_chat = new_name

# -------------------
# Streamlit UI
# -------------------
st.title(f"📊 {st.session_state.current_chat}")
st.subheader("📰 Latest Financial News")

news_articles = get_financial_news()
for art in news_articles:
    st.markdown(f"**{art['title']}**  \n[Read more]({art['link']})  \n*Published: {art['published']}*")
    st.write("---")

user_query = st.text_input("Ask me about finance, stocks, or banking:")

if user_query:
    with st.spinner("Retrieving financial context..."):
        docs, sources = retrieve_docs(user_query)
        context_str = "\n".join(docs)
        user_message = {
            "role": "user",
            "content": f"Answer the query based on the context below:\n\n{context_str}\n\nQuestion: {user_query}"
        }
        st.session_state.chats[st.session_state.current_chat].append(user_message)
        answer = asyncio.run(async_together_chat(st.session_state.chats[st.session_state.current_chat]))
        st.session_state.chats[st.session_state.current_chat].append({"role": "assistant", "content": answer})

    st.write("### 💡 Answer")
    st.write(answer)

    st.write("### 📖 Sources")
    for src in sources:
        st.write(f"- {src}")

# -------------------
# Display chat history
# -------------------
st.subheader("💬 Chat History")
for msg in st.session_state.chats[st.session_state.current_chat]:
    if msg["role"] == "user":
        st.write(f"🧑 **You:** {msg['content']}")
    elif msg["role"] == "assistant":
        st.write(f"🤖 **Bot:** {msg['content']}")