{"id":50606965,"url":"https://github.com/arach/lab","last_synced_at":"2026-06-06T00:02:51.378Z","repository":{"id":349431026,"uuid":"1202318176","full_name":"arach/lab","owner":"arach","description":null,"archived":false,"fork":false,"pushed_at":"2026-04-05T22:42:48.000Z","size":1040,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-04-05T23:28:23.766Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"https://arach.dev/lab","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/arach.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-04-05T22:12:41.000Z","updated_at":"2026-04-05T22:42:53.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/arach/lab","commit_stats":null,"previous_names":["arach/lab"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/arach/lab","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/arach%2Flab","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/arach%2Flab/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/arach%2Flab/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/arach%2Flab/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/arach","download_url":"https://codeload.github.com/arach/lab/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/arach%2Flab/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":33964367,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-05T02:00:06.157Z","response_time":120,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-06-06T00:02:48.715Z","updated_at":"2026-06-06T00:02:51.368Z","avatar_url":"https://github.com/arach.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# training-lab\n\nExperiments in voice interfaces, small models, and narrow eval design.\n\nThe repo has two related threads:\n\n- older dictation-to-structure work for spoken syntax and protocol cleanup\n- newer local-intelligence work around short voice notes, useful titles, and\n  tiny intent extraction\n\nThe current focus is the second one.\n\n## Current benchmark direction\n\nThe clean-slate pack in this repo asks one narrow question:\n\n\u003e Given a short voice note, can a model write a useful title and extract only\n\u003e the clearest action intent?\n\nThat pack currently lives here:\n\n- [`eval/local_intelligence/title_intent_v1/cards.json`](/Users/arach/dev/lab/eval/local_intelligence/title_intent_v1/cards.json)\n- [`eval/local_intelligence/title_intent_v1/TITLE_INTENT_V1_SPEC.md`](/Users/arach/dev/lab/eval/local_intelligence/title_intent_v1/TITLE_INTENT_V1_SPEC.md)\n\nThe benchmark is intentionally small:\n\n- `12` cards\n- tiny intent set: `none | research | email | call | schedule`\n- scoring weighted toward title usefulness and restraint\n\nModel/provider inventory and billing heuristics live in:\n\n- [`MODEL_GUIDE.md`](/Users/arach/dev/lab/MODEL_GUIDE.md)\n\nWhy so small:\n\n- title quality is broadly useful in voice-note products\n- intent extraction is only interesting if the model does not overreach\n- a tiny, auditable pack is easier to trust than a broad pseudo-assistant benchmark\n\n## Repo structure\n\n```text\napp/                Next.js site shell for essays, notes, and benchmark views\nblog/               Longform drafts and daily notes\ncomponents/         Reading UI and benchmark panels\neval/               Benchmark packs, runners, and external calibration work\nlib/                TypeScript ports plus site data helpers\npipeline/           Model artifacts and earlier normalization work\nprocessor/          Canonical Python procedural processor\nscripts/            Experiments, harnesses, and utility scripts\ntraining/           Training corpora, converters, adapters, and notes\n```\n\n## What stays from the older work\n\nThe earlier dictation pipeline is still part of the repo because it matters as\nfoundation work:\n\n- spoken syntax reconstruction\n- procedural token scanning\n- protocol cleanup\n- classifier gates for when to use more model help\n\nThat layer still lives in:\n\n- [`lib/index.ts`](/Users/arach/dev/lab/lib/index.ts)\n- [`processor/procedural.py`](/Users/arach/dev/lab/processor/procedural.py)\n- [`scripts/test-protocol-processor.py`](/Users/arach/dev/lab/scripts/test-protocol-processor.py)\n\nIt is just no longer the active front-door story.\n\n## Local benchmark work\n\nThe broader local-intelligence harness is still here:\n\n- [`eval/local_intelligence/README.md`](/Users/arach/dev/lab/eval/local_intelligence/README.md)\n- [`eval/local_intelligence/v2/README.md`](/Users/arach/dev/lab/eval/local_intelligence/v2/README.md)\n\nBut the current repo direction is to narrow before expanding again.\n\n## External calibration\n\nThese experiments remain useful as side probes, not the main benchmark:\n\n- [`eval/news_summarization/README.md`](/Users/arach/dev/lab/eval/news_summarization/README.md)\n- [`eval/qmsum/README.md`](/Users/arach/dev/lab/eval/qmsum/README.md)\n\n## Quick checks\n\n```bash\n# Run the older procedural processor test harness\npython3 scripts/test-protocol-processor.py\n\n# Build the site\nbun run build\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Farach%2Flab","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Farach%2Flab","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Farach%2Flab/lists"}