dev-env-debugger / inference.py

Commit History

  1. clamp all grader scores to strictly (0,1) exclusive per OpenEnv spec

    Ajjack4
    6c700b3
  2. rewrite inference.py to match sample exactly: IMAGE_NAME, HF_TOKEN first, correct log format

    Ajjack4
    87aa853
  3. default ENV_BASE_URL to localhost:7860 — validator runs container there before inference.py

    Ajjack4
    a07333b
  4. add LOCAL_IMAGE_NAME Docker support, use API_KEY var, debug logging for env/model URLs

    Ajjack4
    5d4432b
  5. read API_KEY env var for validator LiteLLM proxy, keep HF_TOKEN/OPENAI_API_KEY as fallbacks

    Ajjack4
    2f0f7b7
  6. guarantee [START]/[END] per task: wrap OpenAI() constructor, guard client=None

    Ajjack4
    ed395b6

  1. fix log format to key=value, default ENV_BASE_URL to live HF Space

    Ajjack4
    91a4057
  2. never exit non-zero: remove sys.exit(1), script always completes cleanly

    Ajjack4
    22ebfb6
  3. harden inference.py: catch all HTTP errors, retry on reset, guard main()

    Ajjack4
    3bd2b79
  4. accept both HF_TOKEN and OPENAI_API_KEY, default API_BASE_URL and MODEL_NAME

    Ajjack4
    c1ff000
  5. handle missing env vars gracefully in inference.py

    Ajjack4
    d6d0673

  1. add inference.py with structured logs, openai client, required env vars

    Ajjack4
    eb9db3e