Prakya/Queue-Waiting-Time-Optimizer-V3

force-redeploy: 20260412_234942

fix: sync latest inference.py to HF Space

Use START/STEP/END prefixed task logs in inference

Use recognized llm grader type for all 3 tasks

Clamp phase 2 scores

Task validation: keep only score-bearing task END lines

Task validation fix: remove non-task END logs from API paths

Task validation hardening: emit unambiguous END score lines

Fix task validation parser conflict: only emit END for scored tasks

Satisfy task validation: emit 3 graded task blocks with bounded scores

Fix output parsing: emit bracketed [START]/[STEP]/[END] logs

Fix OpenEnv reset/post reliability: run uvicorn in container and harden inference imports

Fix phase-2 bind conflict in subproject: run inference entrypoint and add safe fallback

Harden Phase 2 runtime: bounded LLM calls and graceful fallback

Fix Phase 2 timeout: make inference.py execution finite

Fix Phase 2: Correct OpenAI API calls and add port binding fallback

Satisfy multi-mode checks: scripts entrypoint, openenv-core, main(), uv.lock

Align root scaffold with OpenEnv validate requirements

Add root openenv.yaml for OpenEnv validate check

Harden OpenEnv reset/post endpoints and slim Docker runtime

Switch Space config to Docker so root validator files run

Add root OpenEnv entrypoint and Dockerfile for validator checks

Add compliant inference entrypoint and OpenAI dependency

Make Gradio host dynamic for local and HF Spaces

Clean snapshot for Hugging Face Space