Drop flash-attn install (was broken stub); attn_implementation=sdpa; GPU duration 119->60 (free tier max)

Bump bitsandbytes>=0.46.0 (0.44.1 imports removed triton.ops module)

torch>=2.7 with cu128 wheels (Blackwell sm_120 support); drop triton pin (let torch pull compatible)

Also patch recursive _json_schema_to_python_type for bool schemas

Monkey-patch gradio_client.get_type to accept bool schemas (real fix for api_info crash)

Disable API info endpoint (gradio_client bug with bool additionalProperties)

Pin Chatbot to tuples format with explicit empty value (fixes api_info schema crash)

Bump sdk_version 4.36.1 -> 5.6.0 (fixes gradio_client const-in-schema bug)

Drop spaces pin too (HF forces spaces==0.49.3 in build)

Drop gradio pin; let HF SDK auto-install matching version

Update README.md

Update app.py

Update README.md

Update README.md

Update README.md

Update README.md

app.py overall upgrade + img gen Chroma

Upgrade requirements.txt to recent version

Fix triton attempt1

Fix: No candidate PyTorch version found for ZeroGPU

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update

Update app.py

Update app.py

Update app.py

Update app.py

fix typo

Add placeholder model ID

Update app.py

Update app.py

Revert back fix2

Revert back

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Revert fix5