RohanExploit/Meta-hackathon

fix: route ALL diagnostics to stderr, only [START]/[STEP]/[END] on stdout - validator-proof rewrite

fix: run tasks sequentially to prevent interleaved structured output

Merge pull request #7 from RohanExploit/copilot/fix-output-parsing-issues

Fix HF_TOKEN/OPENAI_API_KEY selection to use explicit None checks

Fix output parsing: guard imports, move [START] before TASKS lookup, add global fallback

Merge pull request #6 from RohanExploit/copilot/fix-output-parsing-error

Refine output parsing fix while preserving default parallelism

Fix structured STEP output format and default sequential execution

Merge pull request #5 from RohanExploit/copilot/fix-structured-output-issue-again

Refactor task normalization and improve fallback diagnostics

Ensure START/END pairing on init failures

Improve fallback diagnostics and unknown-task reporting

Harden task normalization and sanitize reset fallback log

Refine inference fallback error handling and task banner output

Fix inference structured stdout emission fallbacks

Merge pull request #4 from RohanExploit/copilot/fix-structured-output-issue

fix: emit required bracketed structured stdout blocks in inference

fix: add openenv-core>=0.2.0 dependency for judge validation

fix: use raw request parsing for /reset and /step to handle null/empty body from judge

fix: accept null body on POST /reset and /step for OpenEnv judge compatibility

fix: load .env in pre_submission_check, soft-fail missing openenv CLI

fix: comply with hackathon checklist rules (env vars, START/STEP/END logs)

feat: harden environment for Meta/OpenEnv judging standards

feat: Add local execution and parallel evaluation for inference

Auto-load .env file if present in inference.py

Fix api fetch paths for iframe/proxy support on HF Spaces

Fix environment instantiation scope and mode tasks

docs: add verified baseline scores (mean=0.62) from Groq e2e test

perf: tune agent prompt + smart fallback for higher fill rates

fix: lazy pipeline imports in inference.py for automated validation compatibility; update README structure

feat: add NVIDIA-inspired decoupled RAG pipeline

Merge pull request #3 from RohanExploit/copilot/upgrade-system-to-realtime

Upgrade to realtime demand tracking, add SSE stream, fix docs and dashboard score

Initial plan

Merge pull request #2 from RohanExploit/copilot/develop-parallel-web-ui

Add parallel modern dashboard UI at /dashboard with Chart.js charts, XSS fixes

Initial plan

Finalizing RITM environment for Meta Hackathon submission: Fixed reward function, updated README, and cleaned up repository.

docs: rewrite README for hackathon judging

fix: per-step reward correctness, demand elasticity, and repo cleanup

chore: add uv.lock, update openenv dependencies, and fix validator

fix: hackathon compliance - root Dockerfile, HF Spaces metadata, OPENAI_API_KEY priority

Merge pull request #1 from RohanExploit/copilot/implement-real-world-task-simulation

feat: add HF Spaces config and tighten OpenEnv compliance checks

chore: outline plan for OpenEnv compliance fixes

Initial plan

Add realtime server loop and production validation suite

Harden grading, determinism, tests, and realtime UI integration

Align terminal UI actions with multi-channel API

Add Bloomberg-style realtime terminal UI