{"id":51076983,"url":"https://github.com/zozo123/env-rosetta","last_synced_at":"2026-06-23T15:02:07.588Z","repository":{"id":357199123,"uuid":"1235863726","full_name":"zozo123/env-rosetta","owner":"zozo123","description":"Same Wordle env, 4 RL frameworks, 4 parallel islo.dev sandboxes — Rosetta stone for RL env design (built on @adithya-s-k's RL_Envs_101)","archived":false,"fork":false,"pushed_at":"2026-05-11T18:23:37.000Z","size":12427,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-11T20:26:53.125Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/zozo123.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-11T18:19:30.000Z","updated_at":"2026-05-11T18:24:05.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/zozo123/env-rosetta","commit_stats":null,"previous_names":["zozo123/env-rosetta"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/zozo123/env-rosetta","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fenv-rosetta","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fenv-rosetta/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fenv-rosetta/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fenv-rosetta/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/zozo123","download_url":"https://codeload.github.com/zozo123/env-rosetta/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zozo123%2Fenv-rosetta/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34694786,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-23T02:00:07.161Z","response_time":65,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-06-23T15:02:01.060Z","updated_at":"2026-06-23T15:02:07.576Z","avatar_url":"https://github.com/zozo123.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# env-rosetta\n\n**The same Wordle env, four RL frameworks, four parallel [islo.dev](https://islo.dev/) sandboxes.**\n\n\u003e **Project page:** https://zozo123.github.io/env-rosetta-page/\n\u003e\n\u003e **Writeup:** [`docs/POST.md`](./docs/POST.md)\n\u003e\n\u003e **Inspired by:** [@adithya_s_k](https://x.com/adithya_s_k)'s [RL_Envs_101](https://github.com/adithya-s-k/RL_Envs_101) — same env reimplemented across 6 frameworks as a \"Rosetta stone\" for RL env design. This is that Rosetta stone hosted **per-sandbox** on islo.dev instead of one shared HuggingFace Space.\n\u003e\n\u003e **Sibling:** [unity-loop](https://github.com/zozo123/unity-loop) (Claude-vision tournament over Unity WebGL variants on islo).\n\n---\n\n## What this is\n\nFour islo.dev sandboxes. Each runs Adithya's Wordle env in a different RL framework dialect:\n\n| Sandbox | Framework | Server style | Live URL |\n|---|---|---|---|\n| `rosetta-openenv` | [OpenEnv](https://github.com/meta-pytorch/OpenEnv) (Meta) | FastAPI + MCP | `cards/openenv.json#share_url` |\n| `rosetta-ors` | [ORS](https://github.com/openreward/openreward) (Open Reward Standard) | HTTP | `cards/ors.json#share_url` |\n| `rosetta-nemo-gym` | [NeMo Gym](https://github.com/NVIDIA/NeMo-RL) (NVIDIA) | HTTP | `cards/nemo_gym.json#share_url` |\n| (in-process) | [Verifiers](https://github.com/willccbb/verifiers) | In-process (no server) | rollout transcript only |\n\nThe same MCP/HTTP shape — `guess(word)`, `get_history()`, `reset_game()` — exposed by every server. Same Wordle word every time you click in (seeded per session). Different framework idioms behind the curtain.\n\n## Why islo.dev for this\n\nAdithya's reference deployments live on HuggingFace Spaces (one Space per framework). HF Spaces is fine for static demos but you don't get a fresh per-trial sandbox, you can't `ssh` into one, and you can't run parallel rollouts without sharing state. **islo.dev gives you cold per-rollout sandboxes** — same code, but you can spin K parallel workers, each isolated, each a one-line provision.\n\n## Quick start\n\n```bash\nbin/env-rosetta provision    # spawn 4 islo sandboxes, install deps, start each server, share ports\nbin/env-rosetta rollout      # run one LLM rollout against each framework, capture transcripts\nbin/env-rosetta page         # update cards/*.json (which the gh-pages site reads)\n```\n\nOr just look at [`cards/`](./cards) — each framework's metadata, share URL, sample code, and a recorded rollout transcript are versioned there.\n\n## Tier 2 — Jupyter env (the real \"islo replaces E2B\" story)\n\nAdithya's Jupyter agent env uses `e2b-code-interpreter` for real Python code execution. Porting that to islo is a separate workstream — write an `IsloSandbox` class with the same `run_code` interface as `E2BSandbox`, plumb it through `envs/jupyter_env/\u003cframework\u003e/e2b_sandbox.py` in all 4 frameworks. Sketched in [`docs/POST.md#tier-2`](./docs/POST.md#tier-2-jupyter-env--islo-replaces-e2b). The Wordle rosetta in this repo is the foundation; Tier 2 is the killer demo.\n\n## Credits\n\n- Envs and framework adapters — [@adithya-s-k](https://github.com/adithya-s-k/RL_Envs_101)\n- Sandbox infra — [islo.dev](https://islo.dev/)\n- Pattern — [unity-loop](https://github.com/zozo123/unity-loop) · [pokeloop](https://github.com/zozo123/pokeloop) · [meta-harness-on-islo](https://github.com/zozo123/meta-harness-on-islo)\n- Movie — [agentreel](https://github.com/islo-labs/agentreel)\n\nMIT.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fzozo123%2Fenv-rosetta","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fzozo123%2Fenv-rosetta","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fzozo123%2Fenv-rosetta/lists"}