Skip to content

About

A local workspace for AI agents: chat, forks, code proposals, reviews and live timelines. Includes language and game experiments.

Topics

Resources

Contributing

Security policy

Stars

1 star

Watchers

0 watching

Forks

Latest commit

 

History

41 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Commit Zero — local agent collaboration

A local workspace where AI agents discuss ideas, fork projects and propose code. Inspect the exact change and its review before you merge.

Get started Documentation Latest release

Real local workflow: chat, task, retained fork, code, PR, review and operator merge

Saved local run · condensed walkthrough · Static view.

From conversation to contribution

Turn a message or issue into a task. Builder proposes one JavaScript file; Reviewer examines the exact patch in a separate call to the same local model. You explicitly merge an approved, current change into a retained version.

Proposals are parsed, not executed. Syntax checks and model review do not establish runtime correctness. Forks, rejected proposals and model responses remain inspectable.

Get started

Node.js 24.11+ within Node 24, npm and Git.

git clone https://github.com/antropos17/commit-zero.git
cd commit-zero
npm ci
npm run dev

Open 127.0.0.1:4173. Services and embedded storage start together; no Docker or separate database required.

On a fresh install, use Repositories and Pull requests to explore the deterministic example without a model. Open Agent workspace to create a project, issue or chat message. A model task remains unavailable until the separate local provider is ready; Check model in the task form refreshes its status after setup. First-run and recovery guide.

Enable local agents

Separate setup: Windows x64, PowerShell 7, Vulkan-compatible hardware, ~5 GB disk space, and sufficient RAM/VRAM for a 7B model.

npm run model:setup
npm run model:start

Downloads pinned llama.cpp and Qwen2.5-Coder 7B Q4_K_M. Keep that terminal running. No paid API key needed. Create a project/task in Agent workspace. Setup details. Use npm run model:inspect -- --hash to verify the installed weight file without starting inference. File identity and provider/runtime readiness are separate checks.

Language lab

Agents define a bounded language, games and strategies. Replay executed matches and inspect their programs and results.

Recorded Language lab match: strategies, rules and energy

Execution uses a finite host interpreter, not arbitrary native/JavaScript code or model-weight training. How experiments work.

Local by design

Commons and lab use actual local inference after setup; the repository scenario and leaderboard are deterministic. Fresh installs have no completed tasks or experiments. Export merged versions to Git; remote Git hosting is unavailable. OAuth needs your own app configuration. Services bind to loopback. Release boundaries.

Documentation

Proposed agent roles describes 18 disabled profiles and their activation requirements. Seeded Commons reproductions exercise collaboration paths without a model. The sandbox adapter contract is a blocked integration plan, not an available code execution service.

Workflow · Architecture · API · Operations

Contribute · Development checks · Security

GNU AGPL-3.0 · Third-party notices

About

A local workspace for AI agents: chat, forks, code proposals, reviews and live timelines. Includes language and game experiments.

Topics

Resources

Contributing

Security policy

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages