app.py

import re, json, gradio as gr
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# LIGHTWEIGHT, CPU-FRIENDLY MODEL
MODEL_ID = "google/mt5-small"
LABELS = ["PERSON","ORG","LOC","GPE","DATE","TIME","PRODUCT","EVENT"]

def build_prompt(text, labels=LABELS):
    return (
        "متن زیر را برای شناسایی موجودیت‌های نامدار (NER) تحلیل کن.\n"
        f"لیبل‌های مجاز: {', '.join(labels)}.\n"
        "خروجی را فقط به صورت JSON معتبر با اسکیمای زیر بده:\n"
        '{"entities":[{"text":"...", "label":"ORG|PERSON|...", "start":0, "end":0}]}\n'
        "هیچ متن دیگری ننویس؛ فقط JSON.\n\n"
        f"متن: {text}\n"
    )

def extract_json(s: str):
    m = re.search(r"\{[\s\S]*\}", s)
    if not m: return {"entities": []}
    raw = m.group(0)
    try:
        return json.loads(raw)
    except Exception:
        raw = re.sub(r",\s*}", "}", raw)
        raw = re.sub(r",\s*]", "]", raw)
        try:
            return json.loads(raw)
        except Exception:
            return {"entities": []}

# Lazy load on CPU
_tokenizer = None
_model = None

def load_model():
    global _tokenizer, _model
    if _tokenizer is None or _model is None:
        _tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, use_fast=True)
        _model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_ID)  # CPU by default on Spaces
    return _tokenizer, _model

def ner_infer(text, max_new_tokens=192):
    text = (text or "").strip()
    if not text:
        return {"entities": []}

    tok, model = load_model()
    prompt = build_prompt(text)
    inputs = tok(prompt, return_tensors="pt")  # stays on CPU

    gen_ids = model.generate(
        **inputs,
        max_new_tokens=int(max_new_tokens),
        do_sample=False,          # deterministic on CPU
        temperature=0.0,
        # pad_token_id must be set for some T5/mT5 variants:
        pad_token_id=tok.pad_token_id,
        eos_token_id=tok.eos_token_id
    )
    out = tok.decode(gen_ids[0], skip_special_tokens=True)
    data = extract_json(out)

    # normalize; if model omits start/end, compute first occurrence
    ents = []
    for e in data.get("entities", []):
        try:
            t = str(e.get("text","")).strip()
            lab = str(e.get("label","")).strip()
            if not t or not lab: 
                continue
            st = e.get("start"); en = e.get("end")
            if not isinstance(st, int) or not isinstance(en, int):
                idx = text.find(t)
                if idx >= 0:
                    st, en = idx, idx + len(t)
                else:
                    st, en = 0, 0
            ents.append({"text": t, "label": lab, "start": int(st), "end": int(en)})
        except Exception:
            pass
    return {"entities": ents}

with gr.Blocks(title="Persian Zero-Shot NER (CPU)") as demo:
    gr.Markdown("## Persian Zero-Shot NER (LLM) — CPU version (mT5-small)")
    inp = gr.Textbox(label="متن فارسی", lines=4, value="من دیروز با علی در تهران در دفتر دیجی‌کالا جلسه داشتم.")
    max_tok = gr.Slider(64, 512, value=192, step=16, label="Max new tokens (CPU)")
    btn = gr.Button("Extract Entities")
    out = gr.JSON(label="خروجی JSON")
    btn.click(fn=ner_infer, inputs=[inp, max_tok], outputs=out)

if __name__ == "__main__":
    demo.launch()