[sergiopaniego]

Adds a Sudoku tutorial that follows the Wordle one (#1479): Qwen3-1.7B, an SFT warmup on solver-played games, then GRPO against the TextArena Sudoku-v0 environment, with one place(row, col, number) tool call per cell.

• examples/sudoku/: puzzles, SFT data, SFT, GRPO, evaluation, summary and GIF scripts, plus the 30 fixed evaluation puzzles at 26 and 16 empty cells.
• docs/source/tutorials/sudoku-grpo.md: the whole pipeline with the numbers, hardware and cost of our run (about $42 in total on Hugging Face Jobs).

Results on the same 30 puzzles, greedy:

Model	26 empty cells: solved	cells filled	16 empty cells: solved	cells filled
Qwen3-1.7B	0/30	0.1%	0/30	1.9%
SFT	10/30	55.8%	19/30	79.1%
SFT + GRPO (75 steps)	22/30	83.2%	26/30	87.5%

Models, data and curves: SFT, SFT + GRPO, dataset, trackio.

The "Long runs against a Space" section explains the mid-game reconnects we hit during GRPO, which led to #1536.

The two images point to huggingface/documentation-images (openenv/sudoku-progress.gif and openenv/sudoku-grpo-curve.png).

---

Note

Low Risk
Changes are limited to documentation and standalone example scripts; no core OpenEnv runtime or training library code is modified in this diff.

Overview
Adds an end-to-end Train a Sudoku Agent path parallel to the Wordle tutorial: docs navigation and a long-form guide (sudoku-grpo.md), plus a new examples/sudoku/ package.

The example wires Qwen3-1.7B to TextArena Sudoku-v0 with one place(row, col, number) tool call per cell. Scripts cover puzzle generation, solver-driven SFT dataset export, SFT and GRPO (TRL GRPOTrainer + environment_factory against a duplicated HF Space), greedy evaluation on fixed puzzles_c55.json / puzzles_c65.json, summarization, and a side-by-side GIF helper. GRPO uses dual rewards (solved + cells-filled progress), memory-safe batching for ~16k-token games, and closes/reopens the Space client on every game to avoid mid-episode WebSocket drops corrupting training (documented alongside #1536).

Doc updates link the tutorial from the Training sidebar and the tutorials index table; the guide includes run costs, hardware, and operational notes (Space env vars, Trackio, Hub artifacts).

^{Reviewed by Cursor Bugbot for commit a54055a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit a2dd4ef. Configure here.}