services/summarizer.py

import os
import logging
from openai import OpenAI
from huggingface_hub import InferenceClient

logger = logging.getLogger(__name__)

# Fallback open-source model via HF Serverless API
HF_MODEL = "facebook/bart-large-cnn"
HF_MAX_INPUT_CHARS = 4000  # truncate input to roughly 1000 tokens to avoid timeouts/limits

# OpenAI
OPENAI_MODEL = "gpt-4o-mini"
OPENAI_MAX_INPUT_CHARS = 40000  # larger context window

def generate_summary(text: str, is_registered: bool = False) -> str:
    """
    Generate a summary of the extracted text.
    Uses OpenAI for registered users (if key exists), 
    and Hugging Face Serverless API for anonymous users.
    """
    if not text or len(text.strip()) < 50:
        return "Not enough text to generate a meaningful summary."

    if is_registered:
        api_key = os.environ.get("OPENAI_API_KEY")
        if api_key:
            return _summarize_openai(text, api_key)
        else:
            logger.warning("OPENAI_API_KEY not found. Falling back to HF Serverless API.")
            return _summarize_hf(text)
    else:
        return _summarize_hf(text)

def _summarize_openai(text: str, api_key: str) -> str:
    try:
        client = OpenAI(api_key=api_key)
        truncated_text = text[:OPENAI_MAX_INPUT_CHARS]
        response = client.chat.completions.create(
            model=OPENAI_MODEL,
            messages=[
                {"role": "system", "content": "You are a professional assistant. Provide a concise, highly readable 2-3 sentence summary of the following document. Focus on the core facts, purpose, or conclusions."},
                {"role": "user", "content": truncated_text}
            ],
            max_tokens=150,
            temperature=0.3
        )
        return response.choices[0].message.content.strip()
    except Exception as e:
        logger.error(f"OpenAI summarization failed: {e}")
        return ""

def _summarize_hf(text: str) -> str:
    try:
        # Use InferenceClient (uses the HF_TOKEN env var if present in Space, or anonymous if not)
        client = InferenceClient()
        truncated_text = text[:HF_MAX_INPUT_CHARS]
        response = client.summarization(truncated_text, model=HF_MODEL)
        
        if isinstance(response, list) and len(response) > 0:
            return response[0].get("summary_text", "").strip()
        elif isinstance(response, dict):
            return response.get("summary_text", "").strip()
        else:
            return str(response)
    except Exception as e:
        logger.error(f"HF summarization failed: {e}")
        return ""