[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-82763":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":14,"stars7d":16,"stars30d":17,"stars90d":15,"forks30d":15,"starsTrendScore":18,"compositeScore":19,"rankGlobal":9,"rankLanguage":9,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":23,"createdAt":9,"pushedAt":9,"updatedAt":24,"readmeContent":25,"aiSummary":26,"trendingCount":15,"starSnapshotCount":15,"syncStatus":27,"lastSyncTime":28,"discoverSource":29},82763,"autoform-bot","facebookresearch\u002Fautoform-bot","facebookresearch","Autoform Bot",null,"Python",75,15,52,1,0,7,19,6,51.01,"Other",false,"main",[],"2026-06-12 04:01:38","# Autoformalization Pipeline\n\nMulti-agent system for translating LaTeX mathematics into verified Lean 4 proofs using Mathlib.\n\n![Visualizer Dashboard](docs\u002Fassets\u002Fvisualizer.png)\n\n## Setup\n\n```bash\nmake setup    # creates venv, installs deps, builds Lean + REPL (~20 min)\n```\n\nCreate `.env` with the API key for your chosen provider:\n```\nANTHROPIC_API_KEY=your-key-here   # for Claude models\nOPENAI_API_KEY=your-key-here      # for GPT models\nGEMINI_API_KEY=your-key-here      # for Gemini models\n```\n\n## Quick Start\n\n**1. Prepare book data** — place `book.md` (and optionally `book.pdf`) in `autoform\u002Fdata\u002F\u003Cname>\u002F`. See `autoform\u002Fdata\u002Fexample\u002F` for a sample.\n\n**2. Extract targets:**\n```bash\npython -m autoform.statement_extraction run \\\n    --book-dir=autoform\u002Fdata\u002Fmy_book \\\n    --output=autoform\u002Fdata\u002Fmy_book\u002Ftargets.yaml\n```\n\n**3. Create a config** (see `autoform\u002Fbot\u002Fconfigs\u002F` for examples):\n```yaml\nworkspace:\n  path: ..\u002Fmy-workspace\n  mathlib_path: submodules\u002Fmathlib\n  lib_name: My_Book\n\nbook:\n  path: my_book\n  files: [book.md]\n  targets: targets.yaml\n\nllm:\n  model: Opus 4.6\n\nworkers:\n  agents_per_node: 5\n  num_repls_per_node: 5\n  min_agents_per_task: 3\n  max_agents_per_task: 5\n```\n\n**4. Run:**\n```bash\n# Start fresh\npython -m autoform.bot.main run --config=path\u002Fto\u002Fconfig.yaml --name=my-run --fresh\n\n# Resume an interrupted run (omit --fresh)\npython -m autoform.bot.main run --config=path\u002Fto\u002Fconfig.yaml --name=my-run\n\n# Multi-node with SLURM\nsrun --nodes=N --ntasks-per-node=1 python -m autoform.bot.main run --config=... --name=my-run\n```\n\n**5. Monitor:**\n```bash\npython -m autoform.visualizer.app --runs-dir=..\u002Fmy-workspace --port=8003\n```\n\n**6. Evaluate:**\n```bash\npython -m autoform.eval run \\\n    --repo-dir=..\u002Fmy-workspace\u002Fmy-run\u002Fcode \\\n    --task-file=autoform\u002Fdata\u002Fmy_book\u002Ftargets.yaml \\\n    --book-dir=autoform\u002Fdata\u002Fmy_book\n```\n\n## Architecture\n\n```\nautoform-pipeline\u002F\n├── autoform\u002F\n│   ├── bot\u002F                  Multi-agent pipeline (orchestrator, workers, reviewers)\n│   ├── eval\u002F                 Evaluation (grading, lean checks, metrics, rubrics)\n│   ├── visualizer\u002F           Web dashboard for inspecting runs and traces\n│   ├── statement_extraction\u002F Statement chunking and extraction from LaTeX\n│   └── data\u002F                 Book datasets (book.md + targets.yaml)\n├── core\u002F                     Framework (agent, inference, trace, coordination)\n├── tools\u002F                    MCP tool servers (filesystem, git, bash, Lean REPL\u002FLSP, mathlib)\n├── template\u002F                 Lean 4 + Mathlib workspace template\n├── submodules\u002F               Git submodules (mathlib, repl, lean-lsp-mcp)\n└── docs\u002F                     Documentation\n```\n\n## Documentation\n\n**Pipeline:**\n- [Bot](docs\u002Fpipeline\u002Fbot.md) — multi-agent architecture, DAG workflow, multi-node SLURM, agent roles, config reference\n- [Evaluation](docs\u002Fpipeline\u002Feval.md) — matching, axiom checking, LLM grading rubrics, dependency graphs\n- [Statement Extraction](docs\u002Fpipeline\u002Fstatement_extraction.md) — chunking, multi-agent extraction, deduplication\n- [Visualizer](docs\u002Fpipeline\u002Fvisualizer.md) — dashboard views, API endpoints, hub mode\n\n**Tools:**\n- [Tools Overview](docs\u002Ftools\u002Foverview.md) — MCP tool system, available servers, adding new tools\n- [REPL Reference](docs\u002Ftools\u002Frepl.md) — Lean REPL architecture, pooled server, Python API\n\n## License\n\nThis project is licensed under the [MIT](LICENSE) license.\n\n\n## Citation\n\nIf you find this work useful, please cite our paper:\n\n```bibtex\n@misc{rammal2026formalizingmathematicsscale,\n      title={Formalizing Mathematics at Scale}, \n      author={Ahmad Rammal and Niket Patel and Fabian Gloeckle and Amaury Hayat and Julia Kempe and Remi Munos and Charles Arnal and Vivien Cabannes},\n      year={2026},\n      eprint={2605.29955},\n      archivePrefix={arXiv},\n      primaryClass={cs.AI},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2605.29955}, \n}\n```\n","Autoform Bot 是一个用于将LaTeX数学内容转换为经验证的Lean 4证明的多代理系统。其核心功能包括从LaTeX文档中提取数学语句，并通过与Mathlib集成，利用多种语言模型（如Claude、GPT和Gemini）来生成和验证这些证明。该工具支持分布式计算环境下的高效运行，提供了一个可视化仪表盘以监控处理流程及结果。适用于需要自动化生成或验证形式化数学证明的研究场景，尤其是在数学、计算机科学领域内进行大规模文献分析时能够显著提高工作效率。",2,"2026-06-11 04:09:09","CREATED_QUERY"]