app/rag_app.py

# app/rag_app.py
import os
import httpx
from openai import OpenAI
from .embeddings import EmbeddingManager
from .store import VectorStore


class RAGApp:
    def __init__(self):
        self.embedder = None
        self.vectorstore = None
        self.client = None

        try:
            # Initialize embedder and FAISS vectorstore
            self.embedder = EmbeddingManager()
            self.vectorstore = VectorStore()

            # Load Hugging Face token
            
            api_key = os.getenv("HF_TOKEN")
            if not api_key:
                raise ValueError("HF_TOKEN not found in environment variables")

            # Use Hugging Face OpenAI-compatible router
            self.client = OpenAI(
                base_url="https://router.huggingface.co/v1",
                api_key=api_key,
                http_client=httpx.Client(timeout=60.0),
            )

            print("✅ RAGApp initialized successfully with Hugging Face router.")

        except Exception as e:
            print("❌ RAGApp init error:", e)
            self.client = None

    # 🧠 Add notes and store embeddings
    def add_notes(self, text):
        chunks = [text[i:i + 1000] for i in range(0, len(text), 800)]
        embeddings = self.embedder.generate_embeddings(chunks)
        self.vectorstore.add_documents(chunks, embeddings)
        return len(chunks)

    # 💬 Query the system
    def ask(self, query):
        try:
            if not self.client:
                return "Error: API client not initialized."

            # 1️⃣ Create embedding for query
            q_embed = self.embedder.generate_embeddings([query])[0]

            # 2️⃣ Retrieve similar chunks
            docs = self.vectorstore.retrieve_similar_docs(q_embed, top_k=3)
            context = "\n\n".join(docs) if docs else "No context found in notes."

            # 3️⃣ Prepare the conversation
            messages = [
                {
                    "role": "system",
                    "content": (
                        "You are a world-class engineering tutor specializing in Electronics, Embedded Systems, and Programming.\n"
                        "Your responses must be clear, technically accurate, and engaging.\n\n"
                        "### Behavior:\n"
                        "1. If the question is conceptual → explain with clarity and real-world relevance.\n"
                        "2. If it involves code → analyze, correct, and explain fixes.\n"
                        "3. If hardware-related → explain theory + circuit/signal behavior.\n"
                        "4. If theory-based from uploaded notes → connect with practical examples.\n\n"
                        "### Output Style:\n"
                        "- Use Markdown.\n"
                        "- Highlight key terms with bold text.\n"
                        "- Use emojis and structured headings for readability.\n"
                        "- Avoid phrases like 'as an AI model'."
                    ),
                },
                {
                    "role": "user",
                    "content": f"Context:\n{context}\n\nQuestion: {query}\nAnswer clearly and in detail below:",
                },
            ]

            # 4️⃣ Call Hugging Face model via router
            completion = self.client.chat.completions.create(
                model="openai/gpt-oss-20b",  # ✅ correct model route
                messages=messages,
                temperature=0.4,
                max_tokens=800,
            )

            # 5️⃣ Return the model's response
            return completion.choices[0].message.content


        except Exception as e:
            print("❌ Error in ask():", e)
            return f"Error: {e}"