Commit History

  1. fix: bind dummy worker to hidden Gradio button for HF ZeroGPU check and add class 38 to LabelRegistry

    f56867f

  1. fix(security): handle query parameters in HMAC signature and default API key fallback

    790c991
  2. feat: pre-download all 8 AI models on boot via huggingface_hub SDK

    5fc4228
  3. fix: use hf_hub_download SDK and validate >1MB binary size to prevent text-pointer segfaults

    04d56e6
  4. fix: mount Gradio at root / and restore background model download & warmup task

    bdc8261
  5. fix: report status healthy when models_ready is True for Hugging Face health check probe

    38d3256
  6. fix: pre-warm models synchronously during lifespan startup to guarantee zero-downtime container readiness

    d055536
  7. fix: harden health check exception handling and fix cnn_loaded state calculation

    69af453
  8. fix: use streamed chunk download with User-Agent header for HF Hub model binaries

    3982649
  9. fix: add safe fallback to yolo_service for missing ONNX models during warmup

    d2d1629
  10. perf: prioritize fast and limited model download and warmup on boot

    bc9c615
  11. fix: use direct HF resolve download to guarantee binary model files without pointer file corruption

    29ea819
  12. perf: optimize model_loader manifest to download essential ONNX models for rapid boot

    34ce6c6
  13. perf: refine warmup to pre-warm primary production models within RAM budget

    9f53731
  14. fix: export environment variables in app/__init__.py for module resolution

    66c5614
  15. perf: reuse existing ONNX sessions across modes pointing to identical model paths

    ad22f50
  16. fix: enforce explicit CPUExecutionProvider to prevent CUDA device discovery segfaults on HF CPU space

    0c272c4
  17. fix: add 1.0s timeout to DB and Redis checks in health endpoint to prevent proxy 503 timeouts

    cd69f93
  18. fix: set relative data/ model default paths in config.py for portable environment execution

    379216b
  19. fix: remove non-ASCII unicode characters from print calls in model_loader.py

    22edca0
  20. perf: non-blocking async model warmup for instant boot and 200 OK health check

    8563384
  21. fix: add root GET / health response route for HF ingress proxy

    b8299ee
  22. fix: safe redis connection check and graceful fallback

    586a69a
  23. fix: default rate_limiter to memory:// when local redis is used

    c229cda
  24. feat: launch uvicorn with full mounted app and add /api/v1/health route

    5b75e17
  25. fix: in-memory fallback for rate limiter when redis is unreachable

    827e99a
  26. feat: migrate full production detection service architecture with 8-model hub sync

    119c847