Commit History
-
Clean up unused files and update manifests
Tarun +1: Claude Sonnet 4.6Deleted: - agent/ (old local torch model, unused) - baseline/ (superseded subprocess runner) - backend/ (empty module) - db/ (Supabase integration, nothing calls it) - benchmark.py (stale: AsyncInferenceClient + broken task routing) - debug_run.py, test.py (dev scripts with outdated API) - output.txt, output2.txt, error.txt, results.json, build.log, upload-help.txt Updated: - openenv.yaml: v2.0.0 — 10 scenarios per tier, max_steps 5, correct team enum, per-phase reward breakdown, updated issue_type enum - requirements.txt: remove supabase, huggingface-hub - pyproject.toml: remove supabase, huggingface-hub from dependencies Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>26e181c
-
Fix Phase 2: tasks with graders + missing runtime endpoints
Tarun +1: Claude Sonnet 4.6- openenv.yaml: add grader field to all 3 tasks (easy/medium/hard) pointing to openenv.graders.grader.GraderEngine - app.py + server/app.py: add required runtime endpoints GET /health, GET /metadata, GET /schema, POST /mcp - app.py + server/app.py: fix main() to be a proper callable function so openenv validate passes (checks for def main() + main() call) - pyproject.toml: remove torch/transformers/accelerate (not used by inference.py or app.py) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>6efd27d -
Clamp reward scores to [0.002, 0.998] instead of [0.0, 1.0]
Tarun +1: Claude Sonnet 4.6Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>fa17f30
-
phase 1 validation v1
Taruna23653a