omdsh-dev/dsh-gomoku
In DSH, you can play backgammon with the AI. You can also let the AI play against each other to see which AI is stronger.
dsh-gomoku is a gomoku (five-in-a-row) plugin for DeepSeek Harness: it adds a 15×15 board to the DSH sidebar so you can play against DeepSeek or any configured model — with no search algorithm or heuristic pruning. Every move comes from pure LLM reasoning, making it a direct test of a model's thinking ability. Three modes: play black, play white, or AI-vs-AI (two different models on the same board). Each side has its own independently configured model and thinking level (Off / High / Max); unsupported levels auto-fall-back. Per-side system prompts can be edited independently — constrain the AI's chain-of-thought before every move and watch its strength evolve under your 'training'. Per-side reasoning logs collapse beside the board and expand with a click. Pause button interrupts AI thinking at any time; while paused you can place stones manually for both sides; Resume restarts AI. Fixed per-move end-to-end timeout. The board panel opens and closes freely while the game and AI thinking continue in the background — get real work done in DSH and squeeze in a move on the side. Free-style (no forbidden moves) rules on a 15×15 board: double-threes, double-fours, and overlines are all legal; five or more in a row wins.
Install
dsh plugin --profile web add @yejiming/dsh-gomokunpm package @yejiming/dsh-gomoku 0.0.1 (registry-verified 2026-08-23). Install from npm (recommended, prebuilt): dsh plugin --profile web add @yejiming/dsh-gomoku. Install from GitHub source (lib/ is committed, no build at install time, no allowBuilds needed): dsh plugin --profile web add github:omdsh-dev/dsh-gomoku. Local tarball: pnpm pack in the repo, then dsh plugin --profile web add ./yejiming-dsh-gomoku-0.0.1.tgz. All three install methods consume the committed prebuilt lib/ artifacts — no build step runs at install time.
Compatibility
DeepSeek Harness web profile. The board window is a DSH plugin panel (sidebar); the game and AI thinking continue in the background even when the panel is closed. No search algorithm or heuristic pruning — every move comes from pure LLM reasoning via the configured model. Both players' model and thinking level (Off / High / Max) are configured independently; unsupported thinking levels automatically fall back to the model's default. Fixed end-to-end timeout per move with explicit reasoning logs per side (collapsed by default, click to expand). Prebuilt lib/ is committed — no build tooling or allowBuilds required for any install method.
Details
- Repo: omdsh-dev/dsh-gomoku
- Category: Entertainment & Lifestyle
- Stars: 14
- Version: npm package @yejiming/dsh-gomoku 0.0.1 (registry-verified 2026-08-23)
- Last push: 2026-08-15
- First seen: 2026-08-08
Recent updates
The current English README (README.en.md, primary English) documents: three game modes (play black, play white, AI-vs-AI), per-side model and thinking level configuration, per-side system prompt editing, per-side reasoning logs, Pause/Resume, fixed timeout, free-style 15×15 rules, background-continuation while panel closed, three install methods (npm / GitHub / tarball) all using prebuilt lib/ with no build step.
FAQ
- Can the AI actually play a reasonable game of gomoku?
- Yes — every move is generated by the configured model's pure LLM reasoning. The AI-vs-AI mode is particularly revealing: two models compete and sharper reasoning wins. You can edit the system prompt for each side to influence its chain-of-thought and see strength change in real time.
- Does installing from GitHub require a build step?
- No — the prebuilt lib/ output is committed to the repo. All three install methods (npm, GitHub, local tarball) consume the prebuilt artifacts; no build tooling or allowBuilds entry is needed.
- Can I pause the AI mid-game to place moves manually?
- Yes — the Pause button (below the board, next to New Game) interrupts AI thinking at any time. While paused you can place stones for both sides freely with no move limit. Click Resume to restart AI thinking.
Alternatives
omdsh-dev/dsh-annotation · omdsh-dev/dsh-advisor · FlytoMAYDAY80/dsh-pet