main.py

import os
os.environ['HF_HOME'] = '/tmp'

from flask import Flask, request, jsonify, render_template
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
import math, textwrap

app = Flask(__name__)

MODEL_NAME = "sshleifer/distilbart-cnn-12-6"
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME)
summarizer = pipeline("summarization", model=model, tokenizer=tokenizer, device=-1)  # set device appropriately

# Simple mapping of presets to generation lengths
LENGTH_PRESETS = {
    "short": {"min_length": 20, "max_length": 60},
    "medium": {"min_length": 60, "max_length": 130},
    "long": {"min_length": 130, "max_length": 300},
}

def chunk_text_by_chars(text, max_chars=1500, overlap=200):
    if len(text) <= max_chars:
        return [text]
    parts = []
    start = 0
    while start < len(text):
        end = min(len(text), start + max_chars)
        chunk = text[start:end]
        nl = chunk.rfind('\n')
        if nl > max_chars * 0.6:
            end = start + nl
            chunk = text[start:end]
        parts.append(chunk.strip())
        start = end - overlap
    return parts

def apply_tone_instruction(text, tone):
    tone = (tone or "neutral").lower()
    if tone == "formal":
        instr = "Summarize in a formal, professional tone:"
    elif tone == "casual":
        instr = "Summarize in a casual, conversational tone:"
    elif tone == "bullet":
        instr = "Summarize into short bullet points:"
    else:
        instr = "Summarize:"
    return f"{instr}\n\n{text}"

# NEW: Route to show summarizer.html (fixes 404)
@app.route("/")
def home():
    return render_template("index.html")

@app.route("/summarize", methods=["POST"])
def summarize_route():
    data = request.get_json(force=True)
    text = data.get("text", "")[:20000]
    length = data.get("length", "medium")
    tone = data.get("tone", "neutral")

    if not text or len(text.split()) < 5:
        return jsonify({"error": "Input too short."}), 400

    preset = LENGTH_PRESETS.get(length, LENGTH_PRESETS["medium"])
    chunks = chunk_text_by_chars(text, max_chars=1500, overlap=200)
    summaries = []

    for chunk in chunks:
        prompted = apply_tone_instruction(chunk, tone)
        out = summarizer(
            prompted,
            min_length=preset["min_length"],
            max_length=preset["max_length"],
            truncation=True
        )[0]["summary_text"]
        summaries.append(out.strip())

    if len(summaries) == 1:
        final = summaries[0]
    else:
        combined = "\n\n".join(summaries)
        prompted = apply_tone_instruction(combined, tone)
        final = summarizer(
            prompted,
            min_length=preset["min_length"],
            max_length=preset["max_length"],
            truncation=True
        )[0]["summary_text"]

    if tone == "bullet":
        lines = [l.strip() for s in final.splitlines() for l in s.split(". ") if l.strip()]
        final = "\n".join(f"- {l.rstrip('.')}" for l in lines[:20])

    return jsonify({"summary": final})

if __name__ == "__main__":
    app.run(debug=True, port=7860)