app.py

import os
import json
import time
import pickle
from langchain_community.vectorstores import FAISS
from langchain.text_splitter import CharacterTextSplitter
from langchain_huggingface import HuggingFaceEmbeddings
from langchain.docstore.document import Document
from flask import Flask, request, jsonify, render_template
import requests
from dotenv import load_dotenv

# Load environment variables from .env file
load_dotenv()

# Get API key from environment variable (HF Spaces sets this)
API_KEY = os.getenv("OPENROUTER_API_KEY")
if not API_KEY:
    raise RuntimeError("OPENROUTER_API_KEY environment variable not set")

HEADERS = {
    "Authorization": f"Bearer {API_KEY}",
    "Content-Type": "application/json",
    "HTTP-Referer": "https://github.com/your-username/KairoAPI",  # Replace with your actual repository
    "X-Title": "KairoAPI ChatBot",
    "OpenRouter-Bypass-Key": API_KEY  # Add bypass key to avoid prompt logging
}

# Ensure data directory exists
DATA_DIR = os.path.join(os.getcwd(), "data")
os.makedirs(DATA_DIR, exist_ok=True)

# Load or preprocess documents
docs_path = os.path.join(DATA_DIR, "preprocessed_docs.pkl")
if os.path.exists(docs_path):
    with open(docs_path, "rb") as f:
        documents = pickle.load(f)
else:
    with open("Chatbotcontent.txt", "r", encoding="utf-8") as file:
        content = file.read()
    text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    docs = text_splitter.split_text(content)
    documents = [Document(page_content=doc) for doc in docs]
    with open(docs_path, "wb") as f:
        pickle.dump(documents, f)

# Initialize the embedding model
print("Initializing embedding model...")
model_cache = os.path.join(DATA_DIR, "model_cache")
os.makedirs(model_cache, exist_ok=True)
embedding_model = HuggingFaceEmbeddings(
    model_name="intfloat/multilingual-e5-small",
    cache_folder=model_cache
)
print("Embedding model initialized successfully.")

# Initialize FAISS index
print("Initializing FAISS index...")
faiss_path = os.path.join(DATA_DIR, "faiss_index")

def create_faiss_index():
    """Create a new FAISS index from documents."""
    try:
        print("Creating new FAISS index...")
        if not os.path.exists(faiss_path):
            os.makedirs(faiss_path, exist_ok=True)
        
        # Create and save the index
        vs = FAISS.from_documents(documents, embedding_model)
        vs.save_local(faiss_path)
        print(f"FAISS index created successfully at {faiss_path}")
        return vs
    except Exception as e:
        print(f"Error creating FAISS index: {str(e)}")
        raise

# Always create a new index in Hugging Face environment
print("Creating fresh FAISS index...")
vectorstore = create_faiss_index()
print("FAISS initialization complete.")

# System prompt
SYSTEM_PROMPT = """<|system|>Reasoning: ON. You are a helpful assistant representing our organization.
Always answer in a clean, first-person tone like a knowledgeable member of the team.
Avoid explaining reasoning steps, just give direct answers.<|end|>
"""

conversation_history = [
    {"role": "system", "content": SYSTEM_PROMPT}
]

# Flask App
app = Flask(__name__)

@app.route("/")
def index():
    return render_template("index.html")

@app.route("/chat", methods=["POST"])
def chat():
    user_input = request.json.get("message")
    if not user_input:
        return jsonify({"error": "No input provided."}), 400

    docs = vectorstore.similarity_search(user_input, k=3)
    context = "\n\n".join([doc.page_content for doc in docs])

    user_prompt = f"""Using the following context, answer the user's question in a clean and natural tone like you're part of the organization. Avoid showing reasoning steps. Be brief but informative.

<context>
{context}
</context>

Question: {user_input}
"""
    conversation_history.append({"role": "user", "content": user_prompt})

    payload = {
        "model": "openai/gpt-3.5-turbo",  # Using a more reliable model
        "temperature": 0.7,
        "top_p": 0.95,
        "max_tokens": 1024,
        "stream": False,
        "messages": conversation_history[-8:],
        "transforms": ["no-tokens"]  # Disable token logging
    }

    try:
        response = requests.post("https://openrouter.ai/api/v1/chat/completions", headers=HEADERS, data=json.dumps(payload), timeout=10)
        if response.status_code == 200:
            reply = response.json()["choices"][0]["message"]["content"]
            conversation_history.append({"role": "assistant", "content": reply})
            return jsonify({"response": reply})
        else:
            return jsonify({"error": f"Error {response.status_code}: {response.text}"}), 500
    except requests.exceptions.Timeout:
        return jsonify({"error": "Request timed out (10s limit exceeded)"}), 504
    except Exception as e:
        return jsonify({"error": f"Unexpected error: {str(e)}"}), 500

if __name__ == "__main__":
    app.run(host="0.0.0.0", port=7860)