[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-74272":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":8,"htmlUrl":8,"language":9,"languages":8,"totalLinesOfCode":8,"stars":10,"forks":11,"watchers":12,"openIssues":13,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":15,"stars7d":16,"stars30d":17,"stars90d":14,"forks30d":14,"starsTrendScore":18,"compositeScore":19,"rankGlobal":8,"rankLanguage":8,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":21,"topics":24,"createdAt":8,"pushedAt":8,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":14,"starSnapshotCount":14,"syncStatus":13,"lastSyncTime":28,"discoverSource":29},74272,"OpenGauss","math-inc\u002FOpenGauss","math-inc",null,"Python",1222,109,16,2,0,6,8,22,18,69.32,"MIT License",false,"main",true,[],"2026-06-12 04:01:14","\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fmorph.new\u002Fopengauss-0-2-2\">\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FOpen%20in-Morph-f23f42?style=for-the-badge\" alt=\"Open in Morph\">\n  \u003C\u002Fa>\n\u003C\u002Fp>\n\n# Open Gauss\n\nOpen Gauss is a project-scoped Lean workflow orchestrator from Math, Inc. It gives `gauss` a multi-agent frontend for the `lean4-skills` `prove`, `draft`, `review`, `checkpoint`, `refactor`, `golf`, `autoprove`, `formalize`, and `autoformalize` workflows, while staging the Lean tooling, MCP\u002FLSP wiring, and backend session state those workflows need.\n\n> New to OpenGauss, Lean tooling, or AI-agent workflows?\n> Start with the [Start Here guide](website\u002Fdocs\u002Fgetting-started\u002Fstart-here.md).\n>\n> Very short version:\n> - Morph: open [https:\u002F\u002Fmorph.new\u002Fopengauss-0-2-2](https:\u002F\u002Fmorph.new\u002Fopengauss-0-2-2), claim or save the session early if Morph offers it, then use `gauss-open-guide`, `gauss`, `\u002Fchat`, or `\u002Fproject init`.\n> - Local: run `.\u002Fscripts\u002Finstall.sh`, then `gauss-open-guide` or `gauss`, then start with `\u002Fchat`, `\u002Fproject init`, or `\u002Fproject create`.\n\nOpen Gauss handles project detection, managed backend setup, workflow spawning, swarm tracking, and recovery. The proving and formalization behavior still comes from `cameronfreer\u002Flean4-skills`; Gauss exposes it through a Gauss-native CLI and project model.\n\nEach lifted slash command spawns a managed backend child agent in the active project and forwards the same argument tail into the corresponding `lean4-skills` workflow command:\n\n- `\u002Fprove ...` -> `\u002Flean4:prove ...`\n- `\u002Fdraft ...` -> `\u002Flean4:draft ...`\n- `\u002Freview ...` -> `\u002Flean4:review ...`\n- `\u002Fcheckpoint ...` -> `\u002Flean4:checkpoint ...`\n- `\u002Frefactor ...` -> `\u002Flean4:refactor ...`\n- `\u002Fgolf ...` -> `\u002Flean4:golf ...`\n- `\u002Fautoprove ...` -> `\u002Flean4:autoprove ...`\n- `\u002Fformalize ...` -> `\u002Flean4:formalize ...`\n- `\u002Fautoformalize ...` -> `\u002Flean4:autoformalize ...`\n\n## Install\n\nIf you want the fastest pinned release path, [https:\u002F\u002Fmorph.new\u002Fopengauss-0-2-2](https:\u002F\u002Fmorph.new\u002Fopengauss-0-2-2) launches the hosted setup in under 10 seconds. The local installers below are the batteries-included path for your own machine and can take up to 10 minutes.\n\nIf you are not already comfortable with OpenGauss, read the [Start Here guide](website\u002Fdocs\u002Fgetting-started\u002Fstart-here.md) before picking a workflow.\n\n### macOS and Linux\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fmath-inc\u002FOpenGauss.git\ncd OpenGauss\n.\u002Fscripts\u002Finstall.sh\n```\n\nThis is the canonical local install path. It bootstraps the local installer runtime, runs the shared `opengauss` installer flow on your machine, and then auto-attaches to the final `gauss` tmux session when possible.\n\nIt will:\n\n1. Install `uv` if needed\n2. Create a repo-local installer environment\n3. Install or upgrade the local runner\n4. Run the shared `opengauss` installer flow on your machine\n5. Auto-attach you to the local `gauss` tmux session when possible, or print the exact `tmux attach -t gauss` command if not\n\nYou can pass normal template-runner flags through to the alternate script, for example:\n\n```bash\n.\u002Fscripts\u002Finstall.sh --plain\n.\u002Fscripts\u002Finstall.sh --secret OPENAI_API_KEY=...\n.\u002Fscripts\u002Finstall.sh --secret ANTHROPIC_API_KEY=...\n```\n\n### Windows (via WSL2)\n\nOpen Gauss on Windows runs through WSL2 using the same shared installer flow.\n\nFrom PowerShell:\n\n```powershell\n.\\scripts\\install.ps1 -WithWorkspace\n```\n\nThis bootstrap:\n\n1. Starts your WSL distro\n2. Clones or updates `OpenGauss` inside your WSL home directory\n3. Runs `.\u002Fscripts\u002Finstall.sh` there, which executes the shared `opengauss` installer flow inside WSL\n\nIf no WSL distro is initialized yet, the bootstrap will install Ubuntu for you with `wsl --install -d Ubuntu`. If that process drops you into the new Linux shell, type `exit` to return to PowerShell and rerun `.\\scripts\\install.ps1 -WithWorkspace`. Windows may also ask to enable WSL features or restart before you rerun the installer.\n\nIf WSL is not installed yet:\n\n```powershell\nwsl --install -d Ubuntu\n```\n\nYou can also install manually inside WSL:\n\n```bash\nwsl\ngit clone https:\u002F\u002Fgithub.com\u002Fmath-inc\u002FOpenGauss.git ~\u002FOpenGauss\ncd ~\u002FOpenGauss\n.\u002Fscripts\u002Finstall.sh\n```\n\nUse a Linux path such as `~\u002FOpenGauss`, not `\u002Fmnt\u002Fc\u002F...`, for the best performance and terminal behavior.\n## Configuration\n\n### 🖥️ Using Local Models (vLLM)\nIf you prefer to run models locally (e.g., using a local GPU) to save on API costs:\n\n1. **Start your vLLM server** (OpenAI-compatible):\n   ```bash\n   python -m vllm.entrypoints.openai.api_server --model \u003Cmodel_name>\n   ```\n\n2. **Point Gauss at that server** with `gauss setup`, or update `OPENAI_BASE_URL` in `~\u002F.gauss\u002F.env`.\n\n### Updating\n\n```bash\ncd OpenGauss\ngit pull\ngauss update\n```\n\n## Quick start\n\nIf you want the plain-language version first, read the [Start Here guide](website\u002Fdocs\u002Fgetting-started\u002Fstart-here.md).\n\n```\ngauss                         # Launch top-level Gauss\n\u002Fchat                         # Turn on inline onboarding and plain-language chat\n\u002Fproject create ~\u002Fmy-project --template-source \u003Ctemplate-or-git-url>\n\u002Fprove 1+1=2                  # Spawn a proving agent after selecting a project\n\u002Fswarm                        # See running agents\n```\n\nIf you already have a Lean project:\n\n```\ncd ~\u002Fmy-lean-project\ngauss\n\u002Fchat                         # Optional: inline onboarding in the current Gauss session\n\u002Fproject init                 # Register it as a Gauss project\n\u002Fprove                        # Start proving\n```\n\n## The core loop\n\n1. Start the CLI with `gauss`\n2. Create or select the active project with `\u002Fproject`\n3. Launch `\u002Fprove`, `\u002Fdraft`, `\u002Freview`, `\u002Fcheckpoint`, `\u002Frefactor`, `\u002Fgolf`, `\u002Fautoprove`, `\u002Fformalize`, or `\u002Fautoformalize`\n4. Gauss spawns a managed backend child session that runs the corresponding `lean4-skills` workflow command in the active project\n5. Use `\u002Fswarm` to track or reattach to running workflow agents\n\n## Project model\n\nGauss treats Lean work as project-scoped by default. Before launching managed workflows, select the active project once and then let Gauss keep spawning backend child agents inside that project root.\n\n- `\u002Fproject init [path] [--name \u003Cname>]` registers an existing Lean repo as a Gauss project\n- `\u002Fproject convert [path] [--name \u003Cname>]` registers an existing Lean blueprint repo\n- `\u002Fproject create \u003Cpath> [--template-source \u003Csource>] [--name \u003Cname>]` bootstraps a project from a template and registers it\n- `\u002Fproject use [path]` pins the current session to an existing Gauss project\n- `\u002Fproject clear` removes the session override and falls back to ambient project discovery\n\nIf you use `\u002Fproject create` often, set a default template once with\n`gauss.project.template_source` in `~\u002F.gauss\u002Fconfig.yaml` or the\n`GAUSS_BLUEPRINT_TEMPLATE_SOURCE` environment variable.\n\nGauss discovers `.gauss\u002Fproject.yaml` upward from the current working directory, but managed workflow child agents launch from the active project root so the forwarded Lean workflow command always runs in the right project context.\n\n## Workflow commands\n\n- `\u002Fprove [scope or flags]` — spawn a guided proving agent\n- `\u002Fdraft [topic or flags]` — draft Lean declaration skeletons\n- `\u002Freview [scope or flags]` — spawn a read-only Lean review agent\n- `\u002Fcheckpoint [scope or flags]` — spawn the Lean checkpoint workflow\n- `\u002Frefactor [scope or flags]` — spawn a Lean refactor workflow agent\n- `\u002Fgolf [scope or flags]` — spawn a Lean proof golfing workflow agent\n- `\u002Fautoprove [scope or flags]` — spawn an autonomous proving agent\n- `\u002Fformalize [topic or flags]` — spawn an interactive formalization agent\n- `\u002Fautoformalize [topic or flags]` — spawn an autonomous formalization agent\n- `\u002Fswarm` — list running workflow agents\n- `\u002Fswarm attach \u003Ctask-id>` — reattach to a running agent\n- `\u002Fswarm cancel \u003Ctask-id>` — cancel a running agent\n\nInside the interactive CLI, Gauss also rewrites common missing-slash forms like `prove ...`, `review ...`, `checkpoint ...`, and `auto-proof ...` into the corresponding managed workflow commands.\n\n## Managed workflow prerequisites\n\n- `gauss.autoformalize.backend` defaults to `claude-code`\n- Built-in backends: `claude-code`, `codex`\n- `claude` or `codex` installed and authenticated for the backend you select\n- Claude auth can come from either:\n  - the normal `claude auth login` flow \u002F Claude credential files\n  - a saved `ANTHROPIC_API_KEY` in `~\u002F.gauss\u002F.env`\n- If both are present, Gauss defaults to Claude's own local auth and only falls back to `ANTHROPIC_API_KEY` when no Claude credentials are available\n- Override with `gauss.autoformalize.auth_mode` in `~\u002F.gauss\u002Fconfig.yaml`:\n  - `auto` (default): prefer local backend auth, then fall back to saved env\u002FAPI-key auth\n  - `login`: ignore staged API-key auth and let the backend use the normal interactive login flow\n  - `api-key`: force the managed session onto saved env\u002FAPI-key auth\n- `uv` or `uvx` available\n- `ripgrep` (`rg`) available for Lean local search\n- An active Gauss project selected via `.gauss\u002Fproject.yaml`\n\nGauss checks these before launch and tells you exactly what is missing.\n`gauss doctor` also reports the managed-workflow backend, auth mode, `uv` \u002F `lake`\navailability, and whether the current working directory resolves to an active\nGauss project.\n\n---\n\nThis repository was forked from `nousresearch\u002Fhermes-agent`.\n","OpenGauss 是一个由 Math, Inc. 开发的项目范围内的Lean工作流编排器。它为`gauss`提供了一个多代理前端，支持如证明、草稿、审查、检查点、重构、高尔夫、自动证明、形式化和自动形式化等`lean4-skills`工作流程，并且管理这些工作流程所需的Lean工具、MCP\u002FLSP连接以及后端会话状态。该项目特别适合需要在数学证明或形式化验证领域使用自动化代理协助完成复杂任务的场景。通过简单的命令行接口，用户可以轻松启动和管理多个相关的工作流程，从而提高工作效率和准确性。","2026-06-11 03:49:47","high_star"]