ShubhamBuchwani/META-Hackathon

Hackathon compliance: normalize rewards to 0-1, reduce max_steps for runtime, and simplify Dockerfile

Fix Phase 2 validation: clamp rewards, improve health check, and refine openenv.yaml

Fix: Finalize multi-mode packaging with module-based graders and scenario data bundling

Fix: Export openenv.graders as global python entry points and fix python packaging wheel omissions

Fix: Revert openenv.yaml grader format back to filepaths (expected by strict validator)

Fix: Update openenv.yaml to use standard Python module entrypoints for graders and schemas

Fix: Use public ECR mirror for Python base image to bypass Docker Hub rate limits

Fix: Pin Dockerfile to Python 3.11.8 to bypass Hub rate limit and update CMD

Fix: Rename entrypoint to main() in server/app.py for strict validation regex rules

Fix: Multi-mode deployment issues (uv.lock, entrypoint, server/app.py)

Fix: Add pyproject.toml to resolve openenv validation missing file error

Fix: Make JSON body optional for /reset to pass automated validation

Compliance: Mandatory Structured Logging [START]/[STEP]/[END] and Sample script parity

Compliance: Rename to inference.py, use openai client, and implement [START]/[STEP]/[END] logging

Fix: Add missing 'Resource' attributes and correct baseline authorization header

Fix: Robust baseline agent and error handling

Fix: Redirect root URL to API documentation

Initialize project for Hugging Face Spaces