app.py

"""GLiNER 2.5 demo server — identical for every demo.

Serves the static page, answers /api/ready, and POSTs {text, ...} to /api/infer,
which it delegates to model.infer(). All demo-specific logic lives in model.py.

    python app.py            # local dev on :7860
"""

import os
import threading
import time

from flask import Flask, request, send_from_directory

import model

APP_DIR = os.path.dirname(os.path.abspath(__file__))
STATIC = os.path.join(APP_DIR, "static")

_LOCK = threading.Lock()  # one inference at a time: CPU threads beat concurrency

print(f"[boot] {model.DEMO['title']} — mock={model.MOCK}", flush=True)
model.warmup()  # load model / validate fixtures before first request
print("[boot] ready", flush=True)

app = Flask(__name__)


@app.route("/")
def index():
    return send_from_directory(STATIC, "index.html")


@app.route("/<path:name>")
def static_file(name):
    if name in ("brand.css", "strip.js", "favicon.svg"):
        return send_from_directory(STATIC, name)
    return ("not found", 404)


@app.route("/api/config")
def config():
    """Demo content for the page: title, subtitle, example chips. No results."""
    return {k: v for k, v in model.DEMO.items() if k != "internal"}


@app.route("/api/ready")
def ready():
    return {"ready": True}


@app.route("/api/infer", methods=["POST"])
def infer():
    data = request.get_json(silent=True) or {}
    t0 = time.perf_counter()
    with _LOCK:
        out = model.infer(data)
    out["ms"] = round((time.perf_counter() - t0) * 1000)
    return out


if __name__ == "__main__":
    app.run(host="0.0.0.0", port=int(os.environ.get("PORT", 7860)), threaded=True)