app.py

from fastapi import FastAPI, Request
from fastapi.responses import JSONResponse
import gradio as gr
import os
import json
import faiss
import numpy as np
from sentence_transformers import SentenceTransformer
import requests
import nest_asyncio
import uvicorn

# ✅ Load RAG-related files
with open("texts.json", "r", encoding="utf-8") as f:
    texts = json.load(f)

index = faiss.read_index("faiss_index.bin")
embed_model = SentenceTransformer("all-MiniLM-L6-v2")

API_KEY = os.environ.get("OPENROUTER_API_KEY")
MODEL = "qwen/qwq-32b:free"

# ✅ FastAPI app
app = FastAPI()

# ✅ Context Retriever with threshold
def get_context(query, top_k=5, threshold=0.3):
    query_vec = embed_model.encode([query])
    D, I = index.search(np.array(query_vec), top_k)

    # Filter based on similarity distance threshold
    results = [(texts[i], d) for i, d in zip(I[0], D[0]) if d >= threshold]
    if not results:
        return ""
    
    return "\n".join([text for text, _ in results])

# ✅ Chat Function (RAG-aware + fallback)
def chat_fn(message, history):
    headers = {
        "Authorization": f"Bearer {API_KEY}",
        "Content-Type": "application/json"
    }

    context = get_context(message)

    if not context.strip():
        return "❌ Sorry, I can't answer that."

    messages = [
        {
            "role": "system",
            "content": (
                "You are a concise and helpful assistant. Use only the context below to answer. "
                "Respond clearly and briefly. Avoid unnecessary details. If the context doesn't contain the answer, reply: "
                "'Sorry, I can't answer that.'\n\nContext:\n" + context
            )
        }
    ]

    for user, assistant in history:
        messages.append({"role": "user", "content": user})
        messages.append({"role": "assistant", "content": assistant})

    messages.append({"role": "user", "content": message})

    payload = {
        "model": MODEL,
        "messages": messages,
        "max_tokens": 200  # Optional: limit response length
    }

    try:
        response = requests.post("https://openrouter.ai/api/v1/chat/completions", headers=headers, json=payload)
        response.raise_for_status()
        reply = response.json()["choices"][0]["message"]["content"]
    except Exception as e:
        reply = f"❌ Error: {e}"

    return reply

# ✅ FastAPI POST endpoint
@app.post("/chat")
async def chat_api(request: Request):
    body = await request.json()
    message = body.get("message")
    history = body.get("history", [])
    response = chat_fn(message, history)
    return JSONResponse(content={"response": response})

# ✅ Gradio Interface
demo = gr.ChatInterface(
    fn=chat_fn,
    title="CODEX MIRXA KAMRAN",
    description="Chat with AI MODEL trained by Mirxa Kamran",
    theme="soft"
)

# ✅ Mount Gradio at root
app = gr.mount_gradio_app(app, demo, path="/")

# ✅ For local development
if __name__ == "__main__":
    nest_asyncio.apply()
    uvicorn.run(app, host="0.0.0.0", port=7860)