[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-82853":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":13,"subscribersCount":13,"size":13,"stars1d":14,"stars7d":15,"stars30d":16,"stars90d":13,"forks30d":13,"starsTrendScore":17,"compositeScore":18,"rankGlobal":8,"rankLanguage":8,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":8,"pushedAt":8,"updatedAt":24,"readmeContent":25,"aiSummary":26,"trendingCount":13,"starSnapshotCount":13,"syncStatus":12,"lastSyncTime":27,"discoverSource":28},82853,"hermes-multi-agent-workflow","tonbistudio\u002Fhermes-multi-agent-workflow","tonbistudio",null,"Python",270,55,2,0,7,100,152,59,5.24,"MIT License",false,"main",true,[],"2026-06-12 02:04:28","# Hermes Multi-Agent Workflow\n\nA reusable skeleton for an **autonomous, multi-agent triage pipeline** built on\n[Hermes](https:\u002F\u002Fgithub.com\u002FNousResearch\u002Fhermes-agent): a fleet of agents that\n**detects** items from sources, **dedups** them, **scores** them against a rubric,\n**researches** them in parallel, **routes** each to a fulfillment path, pauses at\n**one human approval gate**, then **fulfills and delivers** — all coordinated on a\nsingle Hermes Kanban board.\n\nIt ships pre-wired as a worked example (find pain points AI-agent users hit →\nbuild a fix or make an explainer video), so you can read a complete pipeline and\nthen repoint it at your own domain.\n\n> **This is a template, not a turnkey app.** It runs its unit tests and validates\n> its config out of the box, but going live requires setting up your Hermes\n> install, profiles, auth, and scouts (see `docs\u002F07-runbook.md`). The point is to\n> give you — and your coding agent — a clear, working structure to adapt.\n\n## The idea\n\n```\nsources → intake → dedup → score → research (parallel) → route\n                                                            │\n              ┌──────────────────────┬──────────────────────┤\n            path A                 path B                  shelve\n           (prep)                 (prep)                  (auto)\n              └──────────┬───────────┘\n                   ── HUMAN GATE ──   approve · shelve · modify\n              ┌──────────┴───────────┐\n            fulfill                fulfill\n              └──────────┬───────────┘\n                      deliver\n```\n\nThe shape is fixed; **what flows through it is yours.** Everything domain-specific\nlives in one file, `triage.yaml`.\n\n## Quickstart\n\n```bash\npip install -r requirements.txt          # just PyYAML\npython -m cli.triage validate            # check the example config\npython -m unittest discover -s tests     # 12 tests, all generic\npython -m cli.triage scaffold            # print the Hermes setup plan\n```\n\n## Adapt it to your domain\n\nThe whole adaptation is editing `triage.yaml` + the markdown templates it points\nat. Hand your coding agent **`AGENTS.md`** and ask it to walk you through\n`docs\u002F04-adapting-to-your-domain.md`. In brief:\n\n1. Edit `triage.yaml`: sources, rubric, research lanes, route map, paths, roles.\n2. Edit `paths\u002F` templates (scope rails, deliverable specs, proposal formats).\n3. Edit `skills\u002Ftemplates\u002F` (scout queries + orchestrator notes).\n4. `python -m cli.triage validate`, keep `tests\u002F` green.\n5. Follow `docs\u002F07-runbook.md` to set up profiles and go live.\n\n## Repository layout\n\n```\ntriage.yaml              THE config — your whole pipeline (start here)\nAGENTS.md                Guide for the AI agent adapting this template\nengine\u002F                  Generic engine (rarely edited)\n  config.py              Loads + validates triage.yaml\n  engine.py              TriageEngine — all deterministic step logic\n  scoring.py             Rubric scoring (LLM mode + deterministic mode)\n  routing.py             Classification → path\n  dedup.py               Similarity (token-cosine; embedding-ready)\n  item_vault.py          One markdown file per tracked item\n  kanban_store.py        Writes the Hermes Kanban board\n  intake_parser.py       Parses scout reports\n  frontmatter.py         Stdlib YAML-frontmatter for item files\nproposal_actions.py      Human-gate handler (approve\u002Fshelve\u002Fmodify) — config-driven\npaths\u002F                   Per-path templates you customize\n  rails\u002F   specs\u002F   proposals\u002F\nskills\u002Ftemplates\u002F        Scout + orchestrator SKILL.md templates\ncli\u002Ftriage.py            validate \u002F scaffold \u002F init \u002F install\nscripts\u002Fcost_report.py   Per-item spend for the cost gate\ntests\u002F                   Generic engine tests\ndocs\u002F                    Deep-dive docs (architecture, board, config, adapting, …)\nexamples\u002F                Reference configs\n```\n\n## Documentation\n\n- `docs\u002F01-architecture.md` — fat engine \u002F thin skill; how the pieces fit.\n- `docs\u002F02-the-board.md` — Kanban as the bus; dispatcher; fan-in.\n- `docs\u002F03-config-reference.md` — every `triage.yaml` key.\n- `docs\u002F04-adapting-to-your-domain.md` — the step-by-step adaptation guide.\n- `docs\u002F05-pipeline-stages.md` — each stage, and the gotchas to preserve.\n- `docs\u002F06-security.md` — trust surface, scope rails, safe publishing.\n- `docs\u002F07-runbook.md` — profiles, board, crons, go-live.\n- `examples\u002Fai-agent-pain-points\u002FREFERENCE.md` — full write-up of the reference\n  implementation this template was extracted from.\n\n## Security\n\nThis template runs LLM-authored code and shells out, behind one human gate. Read\n**`SECURITY.md`** and `docs\u002F06-security.md` before deploying — and run the\npre-publish secret-scan checklist before open-sourcing an adapted copy.\n\n## Contributing\n\nSee **`CONTRIBUTING.md`**. The golden rule: keep `engine\u002F` domain-agnostic; new\ndomains go in `triage.yaml`, not the code.\n\n## License\n\nMIT — see `LICENSE`.\n\n## Credits\n\nExtracted and generalized from a working single-machine Hermes pipeline. The\nengine is domain-agnostic; the bundled example reflects its origin.\n","Hermes Multi-Agent Workflow 是一个基于 Hermes 构建的自主多代理分诊流水线框架。其核心功能包括从源检测项目、去重、评分、并行研究、路由到完成路径，并在一个人工审批环节暂停，最终完成并交付任务，所有操作都在单个 Hermes 看板上协调进行。技术特点为预配置了一个完整的示例流程，用户可以轻松阅读和调整以适应自己的领域需求。适用于需要自动化处理多个来源信息，并且希望将部分决策过程交给人类审核的情景，如内容审核、客户服务等。通过简单编辑 `triage.yaml` 文件及相关模板即可快速定制化部署。","2026-06-11 04:09:25","CREATED_QUERY"]