main.py

# from fastapi import FastAPI, Response
# from fastapi.middleware.cors import CORSMiddleware
# from fastapi.staticfiles import StaticFiles
# from fastapi.responses import FileResponse
# from pydantic import BaseModel
# from dotenv import load_dotenv
# import os
# import warnings
# warnings.filterwarnings("ignore")
# from langchain_groq import ChatGroq
# from langchain_community.document_loaders import TextLoader
# from langchain_text_splitters import RecursiveCharacterTextSplitter
# from langchain_huggingface import HuggingFaceEmbeddings
# from langchain_community.vectorstores import Chroma
# from langchain_core.messages import HumanMessage, SystemMessage

# load_dotenv()  # loads GROQ_API_KEY from .env
# os.environ["HUGGINGFACEHUB_API_TOKEN"] = os.getenv("HUGGINGFACEHUB_API_TOKEN", "")
# app = FastAPI()

# # Allow your HTML file to call this API
# app.add_middleware(
#     CORSMiddleware,
#     allow_origins=["*"],
#     allow_methods=["*"],
#     allow_headers=["*"],
# )

# # ── Build RAG retriever on startup ──
# print("Loading knowledge base...")
# loader = TextLoader("quantum_entanglement.txt")
# documents = loader.load()

# splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
# chunks = splitter.split_documents(documents)

# embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")
# vectorstore = Chroma.from_documents(chunks, embeddings)
# retriever = vectorstore.as_retriever(search_kwargs={"k": 3})

# print(f"Knowledge base ready. {len(chunks)} chunks indexed.")

# # ── LLM ──
# llm = ChatGroq(
#     groq_api_key=os.getenv("GROQ_API_KEY"),
#     model="llama-3.1-8b-instant",
#     temperature=0.1,
#     max_tokens=1024
# )

# # ── Request schema ──
# class QueryRequest(BaseModel):
#     query: str
#     top_k: int = 3
#     model: str = "llama-3.1-8b-instant"
#     temperature: float = 0.1

# # ── /chat endpoint ──
# @app.post("/chat")
# def chat(req: QueryRequest):
#     # Retrieve relevant chunks
#     docs = retriever.invoke(req.query)
#     # no chunk labels
#     context = "\n\n".join([d.page_content for d in docs])
#     chunks_preview = [d.page_content[:120] for d in docs]

#     system_prompt = f"""You are QuantumAI, an AI assistant exclusively dedicated to quantum mechanics and quantum information science.

# Your knowledge scope is strictly limited to:
# - Quantum entanglement theory, history, and experimental evidence
# - Bell's theorem, Bell inequalities, and EPR paradox
# - Quantum information science: teleportation, cryptography, and computing
# - Quantum hardware: ion traps, superconducting qubits, photonic systems
# - Decoherence, entanglement entropy, and quantum error correction

# You are NOT permitted to answer questions outside this scope under any circumstances.

# Here is the retrieved knowledge context:
# {context}

# Instructions:
# 1. If the question is factual, answer strictly using the provided context.
# 2. If it is a general quantum mechanics question, use your knowledge.
# 3. If combining both, clearly state what is from the document vs general knowledge.
# 4. If the question is outside your scope, politely refuse.

# Respond clearly. Use **bold** for key terms."""

#     # Swap model/temperature dynamically if changed in frontend
#     dynamic_llm = ChatGroq(
#         groq_api_key=os.getenv("GROQ_API_KEY"),
#         model=req.model,
#         temperature=req.temperature,
#         max_tokens=1024
#     )

#     messages = [
#         SystemMessage(content=system_prompt),
#         HumanMessage(content=req.query)
#     ]

#     response = dynamic_llm.invoke(messages)

#     return {
#         "answer": response.content,
#         "chunks_retrieved": len(docs),
#         "chunks_preview": chunks_preview
#     }

# @app.get("/favicon.ico")
# def favicon():
#     return Response(status_code=204)
    
# # ── Health check ──
# @app.get("/health")
# def health():
#     return {"status": "ok", "model": "llama-3.1-8b-instant"}

# # ── Serve the HTML file ──
# @app.get("/")
# def serve_frontend():
#     return FileResponse("chat.html")


from fastapi import FastAPI, Response
from fastapi.middleware.cors import CORSMiddleware
from fastapi.staticfiles import StaticFiles
from fastapi.responses import FileResponse
from pydantic import BaseModel
from dotenv import load_dotenv
import os
import warnings
warnings.filterwarnings("ignore")
from langchain_groq import ChatGroq
from langchain_community.document_loaders import TextLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.vectorstores import Chroma
from langchain_core.messages import HumanMessage, SystemMessage

load_dotenv()
os.environ["HUGGINGFACEHUB_API_TOKEN"] = os.getenv("HUGGINGFACEHUB_API_TOKEN", "")

app = FastAPI()

# ── CORS ──
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_methods=["*"],
    allow_headers=["*"],
)

# ── Build RAG retriever on startup ──
print("Loading knowledge base...")
loader = TextLoader("quantum_entanglement.txt")
documents = loader.load()

splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
chunks = splitter.split_documents(documents)

embeddings = HuggingFaceEmbeddings(
    model_name="sentence-transformers/paraphrase-MiniLM-L3-v2"
)
vectorstore = Chroma.from_documents(chunks, embeddings)
retriever = vectorstore.as_retriever(search_kwargs={"k": 3})

import re
headings = re.findall(r'^[A-Z][A-Z\s\(\)\/\-]{4,}$', documents[0].page_content, re.MULTILINE)
knowledge_scope = "\n".join(f"- {h.strip().title()}" for h in headings[:20] if h.strip())

print(f"Knowledge base ready. {len(chunks)} chunks indexed.")

# ── LLM ──
llm = ChatGroq(
    groq_api_key=os.getenv("GROQ_API_KEY"),
    model="llama-3.1-8b-instant",
    temperature=0.1,
    max_tokens=1024
)

# ── Request schema ──
class QueryRequest(BaseModel):
    query: str
    top_k: int = 3
    model: str = "llama-3.1-8b-instant"
    temperature: float = 0.1

# ── /chat endpoint ──
@app.post("/chat")
def chat(req: QueryRequest):
    docs = retriever.invoke(req.query)
    context = "\n\n".join([d.page_content for d in docs])
    chunks_preview = [d.page_content[:120] for d in docs]

    system_prompt = f"""You are QuantumAI, an AI assistant exclusively dedicated to quantum mechanics and quantum information science.

Your knowledge base covers the following topics:
{knowledge_scope}

You are NOT permitted to answer questions completely unrelated to quantum physics (e.g. cooking, geography, sports).

Here is the retrieved context for this query:
{context}

Instructions:
1. If the question is covered in the retrieved context, use it as your primary source.
2. If it is a general quantum mechanics question not in the context, use your own knowledge.
3. If combining both, clearly state what comes from the document vs general knowledge.
4. Only refuse if the question is genuinely outside quantum physics entirely.

Respond clearly and in depth. Use **bold** for key terms."""

    dynamic_llm = ChatGroq(
        groq_api_key=os.getenv("GROQ_API_KEY"),
        model=req.model,
        temperature=req.temperature,
        max_tokens=1024
    )

    messages = [
        SystemMessage(content=system_prompt),
        HumanMessage(content=req.query)
    ]

    response = dynamic_llm.invoke(messages)

    return {
        "answer": response.content,
        "chunks_retrieved": len(docs),
        "chunks_preview": chunks_preview
    }

@app.get("/favicon.ico")
def favicon():
    return Response(status_code=204)

# ── Health check ──
@app.get("/health")
def health():
    return {"status": "ok", "model": "llama-3.1-8b-instant"}

# ── Serve React build (must come AFTER API routes) ──
REACT_BUILD = os.path.join(os.path.dirname(__file__), "frontend", "build")

if os.path.exists(REACT_BUILD):
    # Serve static assets (JS, CSS, images)
    app.mount("/static", StaticFiles(directory=os.path.join(REACT_BUILD, "static")), name="static")

    @app.get("/")
    def serve_index():
        return FileResponse(os.path.join(REACT_BUILD, "index.html"))

    @app.get("/{full_path:path}")
    def serve_spa(full_path: str):
        """Catch-all: serve React's index.html for any unknown route (SPA routing)."""
        file_path = os.path.join(REACT_BUILD, full_path)
        if os.path.isfile(file_path):
            return FileResponse(file_path)
        return FileResponse(os.path.join(REACT_BUILD, "index.html"))
else:
    # Fallback: warn that the React build hasn't been generated yet
    @app.get("/")
    def serve_no_build():
        return Response(
            content="<h2>React build not found. Run <code>cd frontend && npm install && npm run build</code> first.</h2>",
            media_type="text/html",
            status_code=503,
        )