Delete streamlit_app.py

streamlit_app.py

@@ -1,216 +0,0 @@
-import streamlit as st
-import os
-from dotenv import load_dotenv
-
-st.set_page_config(page_title="🚀 NVIDIA AI Assistant", page_icon="📈", layout="wide")
-
-import pandas as pd
-import joblib
-from datetime import datetime
-# LangChain & LangGraph
-from langchain_openai import ChatOpenAI # type: ignore
-from langchain_chroma import Chroma # type: ignore
-from langchain_huggingface import HuggingFaceEmbeddings # type: ignore
-from langchain_community.tools import DuckDuckGoSearchRun # type: ignore
-from langchain_core.messages import SystemMessage, HumanMessage # type: ignore
-from langgraph.graph import StateGraph, END # type: ignore
-from typing import TypedDict, Annotated
-import operator
-
-DRIVE_DB_PATH = "./chroma_db_v2"
-CSV_PATH = "nvda_2014_to_2026.csv"
-MODEL_PATH = "nvidia_price_model.pkl"
-
-OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
-if not OPENAI_API_KEY:
-    st.error("❌ OPENAI_API_KEY not found in .env")
-    st.stop()
-
-llm = ChatOpenAI(model="gpt-5.4", temperature=0.3, max_tokens=1024)
-
-# ======================== LOAD COMPONENTS ========================
-@st.cache_resource
-def load_rag():
-    try:
-        device = "cuda" if os.system("nvidia-smi > nul 2>&1") == 0 else "cpu"
-        embeddings = HuggingFaceEmbeddings(model_name="all-mpnet-base-v2", model_kwargs={"device": device})
-        import chromadb # type: ignore
-        from chromadb.config import Settings # type: ignore
-        client = chromadb.PersistentClient(path=DRIVE_DB_PATH, settings=Settings(allow_reset=True))
-        vectorstore = Chroma(client=client, collection_name="nvidia_annual_reports_2014_2025", embedding_function=embeddings)
-        st.sidebar.success("✅ RAG Loaded")
-        return vectorstore
-    except Exception as e:
-        st.sidebar.warning(f"RAG issue: {e}")
-        return None
-
-vectorstore = load_rag()
-
-df = pd.read_csv(CSV_PATH, skiprows=[1])
-df['Date'] = pd.to_datetime(df['Date'])
-df = df.sort_values('Date').reset_index(drop=True)
-
-ddg_search = DuckDuckGoSearchRun()
-
-# ======================== LANGGRAPH STATE ========================
-class AgentState(TypedDict):
-    query: str
-    response: str
-    debug_log: Annotated[str, operator.add]
-    next_node: str
-
-# ======================== NODES ========================
-def router_node(state: AgentState) -> AgentState:
-    q = state["query"].lower()
-    if any(x in q for x in ["predict", "forecast", "price", "7 day", "next week", "tomorrow"]):
-        state["next_node"] = "trader"
-        state["debug_log"] = "🚦 Router → Trader\n"
-    elif any(x in q for x in ["news", "outlook", "latest", "blackwell", "risk", "revenue", "financial"]):
-        state["next_node"] = "researcher"
-        state["debug_log"] = "🚦 Router → Researcher\n"
-    else:
-        state["next_node"] = "general"
-        state["debug_log"] = "🚦 Router → General\n"
-    return state
-
-
-def researcher_node(state: AgentState) -> AgentState:
-    query = state["query"]
-    context = ""
-    if vectorstore:
-        docs = vectorstore.similarity_search(query, k=5)
-        context = "\n\n".join([d.page_content[:700] for d in docs])
-
-    news = ddg_search.run(f"NVIDIA {query} latest news OR earnings")[:800]
-
-    resp = llm.invoke([
-        SystemMessage(content="You are NVIDIA expert researcher."),
-        HumanMessage(content=f"Query: {query}\n\nRAG:\n{context}\n\nNews:\n{news}")
-    ]).content
-
-    state["response"] = resp
-    state["debug_log"] += "🔎 Researcher completed\n"
-    return state
-
-
-def trader_node(state: AgentState) -> AgentState:
-    debug = "📈 Trader (Prophet) activated → "
-    try:
-        checkpoint = joblib.load(MODEL_PATH)
-        model = checkpoint['prophet_model']
-        last_close = checkpoint['last_close']
-        last_date = checkpoint['last_date']
-
-        # Create future dataframe
-        future = model.make_future_dataframe(periods=7, freq='B')
-
-        # Properly add regressors for ALL rows (critical fix)
-        future['MA7'] = df['Close'].rolling(7).mean().iloc[-1]
-        future['MA30'] = df['Close'].rolling(30).mean().iloc[-1]
-        future['Vol7'] = df['Close'].tail(7).std()
-        future['Volume_MA7'] = df['Volume'].tail(7).mean()
-
-        forecast = model.predict(future)
-        pred = forecast.tail(7)[['ds', 'yhat', 'yhat_lower', 'yhat_upper']].copy()
-        pred['ds'] = pred['ds'].dt.strftime('%Y-%m-%d')
-
-        final_pred = pred['yhat'].iloc[-1]
-        upside = (final_pred / last_close - 1) * 100
-
-        if upside > 4.0:
-            trade_rec = f"🟢 **STRONG BUY** – Trade window: next 24-48 hours"
-        elif upside > 1.8:
-            trade_rec = f"🟡 **BUY** – Good entry in next 3 days"
-        elif upside < -2.5:
-            trade_rec = f"🔴 **SELL / CAUTION**"
-        else:
-            trade_rec = "⚪ **HOLD** – Monitor closely"
-
-        lines = [f"{row['ds']}: **${row['yhat']:.2f}** (range ${row['yhat_lower']:.2f}–${row['yhat_upper']:.2f})" 
-                for _, row in pred.iterrows()]
-
-        state["response"] = f"""**🚀 NVIDIA 7-DAY FORECAST**
-
-**Current Price:** ${last_close:.2f} → **Day 7 Expected:** **${final_pred:.2f}** ({upside:+.1f}%)
-
-**Predictions:**
-""" + "\n".join(lines) + f"""
-
-**Trade Recommendation:**
-{trade_rec}
-
-*Backtested MAPE: {checkpoint.get('backtest_mape', 2.23):.2f}%*"""
-
-    except Exception as e:
-        state["response"] = f"Forecast error: {str(e)}"
-
-    state["debug_log"] += debug + "\n"
-    return state
-
-
-def general_node(state: AgentState) -> AgentState:
-    resp = llm.invoke([
-        SystemMessage(content="You are a helpful NVIDIA assistant."),
-        HumanMessage(content=state["query"])
-    ]).content
-    state["response"] = resp
-    state["debug_log"] += "💬 General Agent\n"
-    return state
-
-
-# ======================== BUILD GRAPH ========================
-workflow = StateGraph(AgentState)
-workflow.add_node("router", router_node)
-workflow.add_node("researcher", researcher_node)
-workflow.add_node("trader", trader_node)
-workflow.add_node("general", general_node)
-
-workflow.set_entry_point("router")
-workflow.add_conditional_edges("router", lambda x: x["next_node"],
-    {"researcher": "researcher", "trader": "trader", "general": "general"})
-for node in ["researcher", "trader", "general"]:
-    workflow.add_edge(node, END)
-
-app = workflow.compile()
-
-# ======================== UI ========================
-st.title("🚀 NVIDIA AI Assistant – Multi-Agent System")
-st.caption("NTU DSAI Capstone | RAG + DuckDuckGo + Prophet ML")
-
-if "messages" not in st.session_state:
-    st.session_state.messages = []
-
-for msg in st.session_state.messages:
-    with st.chat_message(msg["role"]):
-        st.markdown(msg["content"])
-
-if prompt := st.chat_input("Ask anything about NVIDIA..."):
-    st.session_state.messages.append({"role": "user", "content": prompt})
-    with st.chat_message("user"):
-        st.markdown(prompt)
-
-    with st.chat_message("assistant"):
-        with st.spinner("Thinking..."):
-            result = app.invoke({
-                "query": prompt,
-                "response": "",
-                "debug_log": "",
-                "next_node": ""
-            })
-            answer = result.get("response", "Sorry, I couldn't generate an answer.")
-            log = result.get("debug_log", "")
-
-            st.markdown(answer)
-            with st.expander("🔍 Debug Trace"):
-                st.code(log)
-
-    st.session_state.messages.append({"role": "assistant", "content": answer})
-
-# Sidebar
-with st.sidebar:
-    st.success("✅ Prophet ML Model Loaded")
-    st.success("✅ RAG Loaded")
-    st.success("✅ DuckDuckGo Ready")
-    if st.button("Clear Chat"):
-        st.session_state.messages = []
-        st.rerun()