Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Fix: switch to Qwen2.5-Coder-32B-Instruct (available on free inference)

Fix: use direct HF Inference API via urllib instead of InferenceClient

Fix: use token param instead of api_key for InferenceClient

Switch to webhook mode: Telegram pushes to us, no outgoing connections needed

Fix: DNS fallback with hardcoded Telegram API IPs for HF Spaces

Rewrite: use raw requests instead of python-telegram-bot to fix DNS issues

Fix: force IPv4 transport to resolve DNS issues on HF Spaces

Fix: add startup delay and retry logic for DNS resolution

Fix: use manual polling lifecycle to avoid set_wakeup_fd thread error

Upload README.md with huggingface_hub

Upload requirements.txt with huggingface_hub

Upload app.py with huggingface_hub

initial commit