tristers-at-elise/hunyuanocr-api

Switch to Docker SDK + vLLM OpenAI server

Switch attn_implementation eager -> sdpa to avoid OOM on full-res pages

Enable show_error=True to surface server-side errors to clients

Drop flash-attn dep (wheel is cp310, Space runs py313; app uses eager attn)

Add /ocr single-shot API endpoint for programmatic access

fix title

fix title

fix title

fix ui

add system role

fix repeat error

debug zerogpu timeout error

debug zerogpu timeout error

debug zerogpu timeout error

debug zerogpu timeout error

debug zerogpu timeout error

debug zerogpu timeout error

debug zerogpu timeout error

debug zerogpu timeout error

fix zerogpu error

fix zerogpu error

fix zerogpu error

fix zerogpu error

fix zerogpu error

fix zerogpu error

fix zerogpu error

fix zerogpu error

update examples

update transformers

init commit

init commit

init commit

init commit

initial commit