{"id":50887770,"url":"https://github.com/kyaukyuai/almanac","last_synced_at":"2026-06-15T18:31:03.087Z","repository":{"id":359997776,"uuid":"1232696033","full_name":"kyaukyuai/almanac","owner":"kyaukyuai","description":"Compile domain-specific, freshness-aware knowledge surfaces as CLI tools, MCP servers, and Claude Code Skills.","archived":false,"fork":false,"pushed_at":"2026-06-08T01:23:53.000Z","size":2293,"stargazers_count":1,"open_issues_count":1,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-08T03:08:10.823Z","etag":null,"topics":["agent-tools","almanac","citations","cli","knowledge-base","llm","mcp","rag","typescript"],"latest_commit_sha":null,"homepage":"","language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/kyaukyuai.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":null,"funding":null,"license":"LICENSE","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":"AGENTS.md","dco":null,"cla":null}},"created_at":"2026-05-08T07:14:39.000Z","updated_at":"2026-06-08T01:23:56.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/kyaukyuai/almanac","commit_stats":null,"previous_names":["kyaukyuai/almanac"],"tags_count":36,"template":false,"template_full_name":null,"purl":"pkg:github/kyaukyuai/almanac","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kyaukyuai%2Falmanac","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kyaukyuai%2Falmanac/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kyaukyuai%2Falmanac/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kyaukyuai%2Falmanac/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/kyaukyuai","download_url":"https://codeload.github.com/kyaukyuai/almanac/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kyaukyuai%2Falmanac/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34376120,"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":"2026-06-15T02:00:07.085Z","response_time":63,"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":["agent-tools","almanac","citations","cli","knowledge-base","llm","mcp","rag","typescript"],"created_at":"2026-06-15T18:31:02.473Z","updated_at":"2026-06-15T18:31:03.069Z","avatar_url":"https://github.com/kyaukyuai.png","language":"TypeScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"# almanac\n\n\u003e Compile domain-specific, freshness-aware knowledge surfaces as CLI tools,\n\u003e MCP servers, and Claude Code Skills.\n\n`almanac` turns a single domain name into a self-contained knowledge surface\nthat host LLMs can use through tools. Each compiled almanac contains sourced\nfacts, freshness metadata, generated tool manifests, runtime contracts, review\nartifacts, and adapters for MCP and Claude Code Skills.\n\nIt is deliberately not a persona or chatbot. The host LLM still does the\nreasoning. `almanac` provides the grounded retrieval layer, live tools, source\ncitations, and validation artifacts.\n\n## What It Builds\n\nGiven a domain such as `sqlite`, `rust`, or `enterprise-ai`, `almanac` compiles:\n\n- a curated source set with accepted and rejected source metadata\n- extracted facts with `fetchedAt`, freshness class, and source citations\n- a SQLite FTS knowledge index, with optional vector retrieval artifacts\n- default tools for facts, official docs, recent web search, and releases\n- 0-3 domain-specific tools generated from the discovered source contract\n- a generic MCP runtime over `listTools`, `execTool`, `listResources`, and\n  `readResource`\n- a Claude Code Skill adapter and client registration support\n- benchmark fixtures and deterministic runtime reports\n- saved run, refresh, and answer artifacts for audits and release gates\n\nThe result can be served by one generic binary:\n\n```bash\nalmanac serve sqlite-demo\nalmanac serve sqlite-demo --transport=http --port=7331\n```\n\n## Requirements\n\n- Bun 1.1.0 or newer\n- `ANTHROPIC_API_KEY` for real LLM-backed compile or answer runs\n- optional `BRAVE_SEARCH_API_KEY` for web source discovery\n- optional `VOYAGE_API_KEY`, `OPENAI_API_KEY`, or `ALMANAC_EMBEDDINGS` for\n  vector retrieval artifacts\n\n## Install\n\n`almanac` is currently source-first and is not published as a package. Clone\nthe repository and install dependencies first:\n\n```bash\ngit clone https://github.com/kyaukyuai/almanac.git\ncd almanac\nbun install\n```\n\nVerify the source entrypoint before linking:\n\n```bash\nbun src/cli.ts --version\nbun src/cli.ts doctor\n```\n\n`doctor` does not call an LLM provider. It reports local setup status and a\nreadiness section for demo, provider-backed compile, answer mode, refresh,\nregistration, and optional judge checks.\n\nThen link the local CLI with Bun:\n\n```bash\nbun link\n```\n\nAfter linking, `almanac` should resolve to this checkout and work from outside\nthe repository:\n\n```bash\ntmpdir=\"$(mktemp -d)\"\n(cd \"$tmpdir\" \u0026\u0026 almanac --version \u0026\u0026 almanac doctor)\n```\n\nFor one-off use without linking, run the same commands with `bun src/cli.ts`\nfrom the repository root.\n\nIf `almanac --version` is stale or points at the wrong checkout, inspect the\nresolved binary and rerun `bun link` from the repository you intend to use:\n\n```bash\nwhich almanac\nreadlink \"$(which almanac)\" 2\u003e/dev/null || true\nbun link\n```\n\n## Quick Start\n\nAsk Almanac for the safest first step. This does not require API keys.\n\n```bash\ntmp=$(mktemp -d)\nalmanac start --root \"$tmp\"\nalmanac start \"Build an almanac for production AI governance checks\" --root \"$tmp\"\nalmanac demo --root \"$tmp\"\nalmanac start --root \"$tmp\"\nalmanac list --root \"$tmp\"\nalmanac status sqlite-demo --root \"$tmp\"\nalmanac inspect sqlite-demo --root \"$tmp\"\nalmanac profile sqlite-demo --root \"$tmp\"\nalmanac benchmark sqlite-demo --root \"$tmp\"\nalmanac ask-fixtures init sqlite-demo --seed-demo --root \"$tmp\"\nalmanac refresh run sqlite-demo \\\n  --from-stage 12-benchmark-run \\\n  --ask-suite \\\n  --save \\\n  --root \"$tmp\"\nalmanac maintain sqlite-demo --dry-run --root \"$tmp\"\nalmanac run sqlite-demo \\\n  --tool query_facts \\\n  --input '{\"q\":\"transactions atomic\"}' \\\n  --root \"$tmp\"\nalmanac export sqlite-demo --root \"$tmp\"\nhandoff_root=$(mktemp -d)\nalmanac import ./almanac-sqlite-demo-0.1.0.tar.gz --root \"$handoff_root\"\nalmanac import ./almanac-sqlite-demo-0.1.0.tar.gz --root \"$handoff_root\" --apply\nalmanac status sqlite-demo --root \"$handoff_root\"\nalmanac wiki sqlite-demo --root \"$tmp\"\n```\n\n`start \"\u003cgoal\u003e\"` is planning-only: it drafts a domain, scope, reference\nchecklist, first questions, and an explicit `almanac new ...` command. It does\nnot compile, call a provider, or write files until you run a follow-up command.\n\nThe demo creates a complete local almanac with curated SQLite facts, source\nreview metadata, default tools, contract files, a Skill adapter, and human\ngolden benchmark fixtures.\n\n`ask-fixtures init --seed-demo` and the Stage 12 `refresh run --ask-suite`\nbootstrap answer readiness without provider credentials for the offline demo.\n`maintain --dry-run` previews future upkeep. `operations` lists explicit\nguided operations, and `studio` opens a local dashboard with Run buttons for\nsafe provider-free operations:\n\n```bash\nalmanac operations sqlite-demo --root \"$tmp\"\nalmanac operations run sqlite-demo \u003coperation-id\u003e --root \"$tmp\"\nalmanac studio --root \"$tmp\"\n```\n\nFor a committed inspection snapshot and answer-mode handoff commands, see\nthe [sample almanacs guide](./docs/sample-almanacs.md).\n\n`export` creates a portable archive that excludes saved `.runs/` records by\ndefault. `import` validates that archive in dry-run mode before writing files;\nadd `--apply` to install it into another root. Use `--include-runs` only when\nthe receiver should get saved tool, refresh, and answer artifacts. `wiki`\ncreates a Markdown inspection bundle for reviewing sources, facts, tools,\nbenchmarks, and the generated file manifest.\n\nIf you did not run `bun link`, replace `almanac` with `bun src/cli.ts` in the\nexamples.\n\n## Compile a Real Almanac\n\nReal compilation uses Anthropic-backed LLM stages for domain analysis, source\ndiscovery, fact extraction, tool design, tool implementation, and benchmark\ngeneration.\n\n```bash\nexport ANTHROPIC_API_KEY=...\nexport BRAVE_SEARCH_API_KEY=... # optional\n\nalmanac new cooking\nalmanac inspect cooking\nalmanac profile cooking\nalmanac sources cooking\nalmanac benchmark cooking\n```\n\nRegister it with a host client:\n\n```bash\nalmanac register cooking --client=claude-code --status\nalmanac register cooking --client=claude-code --apply\nalmanac serve cooking\n```\n\nSupported registration targets are `claude-code`, `claude-desktop`, `cursor`,\nand `codex`. `register --status` is read-only and reports missing Skill files,\nmissing MCP entries, stale installed Skills, and mismatched MCP command paths\nbefore you open the host client.\n\n## Core Commands\n\n| Command | Purpose |\n| --- | --- |\n| `almanac new \u003cdomain\u003e` | Compile an almanac from a domain name. |\n| `almanac demo [id]` | Create a no-key offline demo almanac. |\n| `almanac list` | List installed almanacs with lifecycle status and readiness hints. |\n| `almanac status \u003cid\u003e` | Show whether one installed almanac is usable now and what to do next. |\n| `almanac update \u003cid\u003e --from-stage \u003cstage\u003e` | Re-run part of the compile pipeline. |\n| `almanac feed \u003cid\u003e \u003curl\u003e --apply` | Add one source and reindex without a full rebuild. |\n| `almanac inspect \u003cid\u003e` | Show manifest, stage health, sources, fixtures, and benchmark status. |\n| `almanac profile \u003cid\u003e` | Summarize expertise readiness, evidence, query shapes, and limits. |\n| `almanac sources \u003cid\u003e` | Review accepted and rejected source candidates. |\n| `almanac benchmark \u003cid\u003e` | Run human golden fixtures through the runtime. |\n| `almanac run \u003cid\u003e --tool \u003cname\u003e` | Invoke one compiled tool deterministically. |\n| `almanac ask \u003cid\u003e \u003cquestion\u003e` | Run one LLM-backed cited answer session over compiled tools. |\n| `almanac ask-replay \u003cid\u003e` | Replay saved answer artifacts or JSONL fixtures without an LLM. |\n| `almanac runs \u003cid\u003e` | List, inspect, filter, and prune saved operational artifacts. |\n| `almanac refresh due \u003cid\u003e` | Check refresh readiness without writing files or requiring keys. |\n| `almanac refresh run \u003cid\u003e` | Run a locked manual refresh over the update pipeline. |\n| `almanac serve \u003cid\u003e` | Start the generic MCP server over stdio or Streamable HTTP/SSE. |\n| `almanac register \u003cid\u003e` | Inspect or install Skill and MCP config entries for supported clients. |\n| `almanac export \u003cid\u003e` | Package a compiled almanac as a portable archive. |\n| `almanac import \u003carchive\u003e` | Validate or install an exported archive into a root. |\n| `almanac wiki \u003cid\u003e` | Export a Markdown inspection bundle for review and handoff. |\n| `almanac doctor [id]` | Diagnose local runtime, credentials, root hygiene, artifacts, and readiness. |\n\n## Runtime Model\n\nThe compiled almanac directory is data. The runtime loads that data and exposes\nthe same four-operation contract everywhere:\n\n```ts\ninterface AlmanacRuntime {\n  listTools(): Promise\u003cToolManifest[]\u003e;\n  execTool(name: string, input: unknown): Promise\u003cToolResult\u003cunknown\u003e\u003e;\n  listResources(): Promise\u003cResourceDescriptor[]\u003e;\n  readResource(uri: string): Promise\u003c{ contents: string; mimeType: string }\u003e;\n}\n```\n\nMCP, local CLI execution, answer orchestration, benchmarks, and Skill adapters\nall use this boundary. That keeps the per-domain artifact portable while the\nserving binary stays generic.\n\n## Design Principles\n\n- **Headless runtime + adapters**: MCP and Skill support derive from the same\n  operation contract.\n- **Always fresh**: facts and tools carry freshness policy, TTL, and\n  staleness signals. Stale data is surfaced, not hidden.\n- **Cite or abstain**: tool results must return `citations[]`. No grounded\n  source means no answer.\n- **Compile, do not configure**: source discovery, tool design, code\n  generation, and benchmarks are automated from the domain name, with human\n  review points.\n- **No persona**: an almanac is a retrieval-and-tools layer for a host LLM,\n  not a simulated identity.\n\n## Status\n\n`v0.17.0` is shipped. `v0.18` is in RC validation for Provider-Backed Studio.\nThe 12-stage compile pipeline runs end-to-end against\nmocked and real Anthropic providers, and the runtime is wired into the MCP\necosystem for Claude Code, Claude Desktop, Cursor, and Codex registration.\n\nCurrent shipped lines include:\n\n- `v0.4`: optional vector artifacts, hybrid FTS5/vector RRF retrieval,\n  Streamable HTTP/SSE MCP transport, and wiki exports\n- `v0.5`: deterministic `run --tool`, saved run artifacts, retention cleanup,\n  and portable export hardening\n- `v0.6`: refresh due checks, locked manual refresh runs, refresh artifacts,\n  and scheduler docs\n- `v0.7`: one-shot `ask`, bounded LLM-backed tool planning, cite-or-abstain\n  synthesis, and saved answer artifacts\n- `v0.8`: answer trace diagnostics, deterministic ask replay, answer quality\n  gates, and doctor/profile answer readiness signals\n- `v0.9`: ask fixture authoring, suite-level ask gates, refresh-integrated ask\n  validation, and hardened answer readiness reporting\n- `v0.10`: optional answer entailment judging, compile stability diagnostics,\n  retrieval readiness reporting, and v0.10 RC smoke coverage\n- `v0.11`: source-first install sanity, task-oriented first-run readiness,\n  sqlite-demo handoff, and compile failure recovery UX\n- `v0.12`: installed almanac lifecycle inventory, per-almanac status, import\n  handoff, registration visibility, root hygiene, and cleanup guidance\n- `v0.13`: personal maintenance reports, due-only maintenance apply,\n  repair/cleanup candidates, scheduler handoff snippets, and ask fixture upkeep\n- `v0.14`: guided first-run start, natural-language intake, user-facing\n  vocabulary, no-key answer-readiness bootstrap, and local read-only studio\n- `v0.15`: activation milestones, guided create apply, suggested questions,\n  first-answer guidance, replay prompts, and Studio activation state\n- `v0.16`: guided operation metadata, provider-free operation execution,\n  answer-readiness evidence save, and Studio Run buttons for bounded actions\n- `v0.17`: first-use reports, natural-language setup/source checklist\n  guidance, first-answer promotion, abstention recovery, and Studio first-use\n  visibility\n- `v0.18`: provider readiness reporting, Studio reference/goal intake,\n  confirmation-gated compile and first answer from Studio, and the\n  end-to-end first-use loop without leaving the browser\n\n`v0.17.0` turns guided operations toward the first useful personal almanac:\nempty-root start, natural-language setup planning, source checklist handoff,\nfirst answer promotion, and abstention recovery are surfaced through CLI status\nand Studio before users need to learn the artifact model.\n\n`v0.18` makes Studio provider-backed behind an explicit credential and\nconfirmation boundary: presence-only provider readiness everywhere, reference\nand goal staging in the browser, compile and first answer through the same\ncode paths and artifacts as the CLI, and CLI handoff preserved exactly when\nno provider is detected.\n\nSee [CHANGELOG.md](./CHANGELOG.md) for the concise release history.\n\n## Benchmarks\n\nEach compiled almanac ships with its own generated benchmark fixtures, executed\nend-to-end through the runtime. Latest real-Anthropic smokes at\n`--depth=standard`:\n\n| domain | version | facts | tools (custom) | passed | citationRate | negatives passed |\n| ---: | ---: | ---: | ---: | ---: | ---: | ---: |\n| Enterprise AI | v0.17.0 RC | 173 | 6 (2) | 15/15 | 1.00 | 5/5 |\n| sqlite | v0.3.0 | 620 | 2 | 14/15 | 0.90 | 5/5 |\n| Rust | v0.3.10 | 1438 | 3 | 11/15 | 0.60 | 5/5 |\n\nThe stable signal across the validation runs is that negative fixtures pass:\nout-of-domain or unsupported questions abstain instead of fabricating\ncitations.\n\nThe v0.10.0 Enterprise AI RC smoke also passed a real-provider ask check with a\ncited answer, replayed the saved answer artifact deterministically, promoted it\nto `tests/ask.jsonl`, passed `ask-suite`, persisted refresh-integrated ask\nvalidation, and recorded an optional judge failure with concrete\nunsupported/uncertain claims for review.\n\nThe v0.11.0 no-key RC smoke passed source and linked CLI checks, sqlite-demo\nfirst-run handoff, default export/wiki generation, saved answer replay,\n`ask-suite`, refresh-integrated ask validation, and mocked compile failure\nrecovery on `main`.\n\nThe v0.12.0 RC smoke passed the installed lifecycle gate on `main`: source and\nlinked CLI sanity, sqlite-demo lifecycle status, export/import/wiki handoff,\nregistration visibility with temporary configs, root hygiene checks, dry-run\nremove guidance, Enterprise AI fresh compile benchmark at 305 facts and 15/15\nfixtures, real-provider Enterprise AI ask with two citations, saved-run replay,\nfixture promotion, `ask-suite`, refresh-integrated ask validation, and\ndoctor/profile reporting answer mode as ready.\n\nThe v0.13.0 RC smoke passed the personal maintenance gate on `main`: source and\nlinked CLI sanity, sqlite-demo provider-free maintenance apply, saved\nmaintenance evidence, provider-required boundary checks, repair/cleanup\ndry-runs, scheduler handoff generation for cron/launchd/GitHub Actions,\nEnterprise AI fresh compile benchmark at 407 facts and 15/15 fixtures, valid\nreal-provider Enterprise AI abstention with saved-run replay pass, and\nmaintenance dry-run reporting provider-free planned steps.\n\nThe v0.14.0 RC smoke passed the first-run guided gate on `main`: source and\nlinked CLI sanity, provider-free guided start, natural-language planning,\nsqlite-demo answer-readiness bootstrap via seeded ask fixtures and\nrefresh-backed ask-suite evidence, local read-only studio HTML/API and\nlocalhost boundary checks, Enterprise AI fresh compile benchmark at 185 facts\nand 15/15 fixtures, real-provider Enterprise AI answer with one citation, and\nsaved-run replay pass.\n\nThe v0.15.0 RC smoke passed the guided activation gate on `main`: static gates,\nsource and linked CLI sanity, provider-free activation start, deterministic\nmock guided create apply, sqlite-demo suggested questions, seeded answer\nreadiness, mock first-answer guidance, saved answer replay, Studio activation\nHTML/API and localhost boundary checks, maintenance dry-run, Enterprise AI\nfresh compile benchmark at 138 facts and 15/15 fixtures, valid real-provider\nEnterprise AI abstention with quality pass, and saved-run replay pass.\n\nThe v0.16.0 RC smoke passed the guided operations gate on `main`: static gates,\nsource and linked CLI sanity, sqlite-demo provider-free operation listing,\nCLI operation runner, answer-readiness evidence save, Studio action API/UI,\nlocalhost and POST safety checks, maintenance dry-run, Enterprise AI fresh\ncompile benchmark at 374 facts and 15/15 fixtures, valid real-provider\nEnterprise AI abstention with quality pass, and saved-run replay pass.\n\nThe v0.17.0 RC smoke passed the Guided First Useful Almanac gate on `main`:\nstatic gates, source and linked CLI sanity, empty-root `start`,\nnatural-language first almanac planning, source checklist handoff,\nsqlite-demo provider-free first-answer promotion, abstention recovery, Studio\nfirst-use/recovery visibility, Enterprise AI fresh compile benchmark at 173\nfacts and 15/15 fixtures, valid real-provider Enterprise AI abstention with\nquality pass, and saved-run replay pass.\n\n## Development\n\n```bash\nbun install\nbun run typecheck\nbun test\n```\n\nThe CLI entrypoint is [src/cli.ts](./src/cli.ts). The core runtime surface\nlives in [src/core/runtime.ts](./src/core/runtime.ts), the MCP adapter in\n[src/serve/mcp-server.ts](./src/serve/mcp-server.ts), and the compile stages in\n[src/compile/stages](./src/compile/stages).\n\n## Docs\n\n- [Product guide](./docs/product-guide.md): demo, inspection, source review,\n  benchmarks, refreshes, and saved artifacts\n- [Design document](./docs/design.md): architecture, directory layout,\n  compile stages, and runtime contract\n- [Answer mode](./docs/answer-mode.md): `ask`, citations, abstention, and\n  replay behavior\n- [Sample almanacs](./docs/sample-almanacs.md): committed sqlite-demo wiki\n  snapshot, export handoff, and answer-mode sample commands\n- [v0.9 plan](./docs/v0.9-plan.md): ask fixture authoring, suite gates,\n  refresh validation, and answer readiness hardening\n- [v0.9 RC smoke](./docs/v0.9-rc-smoke.md): release-candidate validation for\n  ask fixture authoring, ask suites, refresh ask validation, and provider ask\n- [v0.10 plan](./docs/v0.10-plan.md): optional entailment judging, compile\n  stability controls, retrieval defaults, and release smoke targets\n- [v0.10 RC smoke](./docs/v0.10-rc-smoke.md): release-candidate validation for\n  entailment judging, compile stability diagnostics, retrieval readiness, and\n  ask-suite workflows\n- [v0.11 plan](./docs/v0.11-plan.md): source-first install, first successful\n  almanac, demo handoff, and compile failure recovery\n- [v0.11 RC smoke](./docs/v0.11-rc-smoke.md): release-candidate validation for\n  first-run readiness, demo handoff, saved answer replay, and failure recovery\n- [v0.12 plan](./docs/v0.12-plan.md): installed almanac lifecycle, inventory,\n  status, import handoff, registration visibility, and cleanup\n- [v0.12 RC smoke](./docs/v0.12-rc-smoke.md): release-candidate validation for\n  installed lifecycle, import handoff, registration visibility, and cleanup\n- [v0.13 plan](./docs/v0.13-plan.md): personal almanac maintenance, maintain\n  runner, repair/cleanup, scheduler handoff, and answer fixture upkeep\n- [v0.13 RC smoke](./docs/v0.13-rc-smoke.md): release-candidate validation for\n  provider-free maintenance, scheduler handoff, and Enterprise AI provider\n  smoke\n- [v0.14 plan](./docs/v0.14-plan.md): first-run guided experience, natural\n  language intake, user-facing terminology, answer readiness bootstrap, and\n  local read-only studio\n- [v0.14 RC smoke](./docs/v0.14-rc-smoke.md): release-candidate validation for\n  guided start, no-key answer readiness bootstrap, local studio, and\n  Enterprise AI provider smoke\n- [v0.15 plan](./docs/v0.15-plan.md): guided activation, explicit create\n  apply, first-answer guidance, activation milestones, and Studio activation\n  surface\n- [v0.15 RC smoke](./docs/v0.15-rc-smoke.md): release-candidate validation\n  for guided activation, mock guided create apply, first-answer guidance,\n  Studio activation, and Enterprise AI provider smoke\n- [v0.16 plan](./docs/v0.16-plan.md): Actionable Studio, guided operations,\n  provider-free local action execution, operation result envelopes, and\n  evidence persistence simplification\n- [v0.16 RC smoke](./docs/v0.16-rc-smoke.md): release-candidate validation for\n  guided operation listing, provider-free operation runner, Studio action\n  API/UI, evidence save, maintenance dry-run, and Enterprise AI provider smoke\n- [v0.17 plan](./docs/v0.17-plan.md): Guided First Useful Almanac, first-use\n  reports, natural-language setup guidance, abstention recovery, and first\n  answer promotion\n- [v0.17 RC smoke](./docs/v0.17-rc-smoke.md): release-candidate validation for\n  empty-root first-use guidance, source checklist handoff, first-answer\n  promotion, abstention recovery, Studio first-use/recovery visibility, and\n  Enterprise AI provider smoke\n- [v0.18 plan](./docs/v0.18-plan.md): Provider-Backed Studio, provider\n  readiness reporting, confirmation-gated compile and first answer, and Studio\n  reference intake\n- [v0.18 RC smoke](./docs/v0.18-rc-smoke.md): release-candidate validation for\n  provider readiness reporting, Studio reference/goal intake,\n  confirmation-gated mock compile and first answer, the end-to-end first-use\n  loop, no-key handoff parity, and Enterprise AI provider smoke\n- [Refresh scheduler](./docs/refresh-scheduler.md): cron, CI, launchd, locks,\n  exit codes, and retention\n- [Changelog](./CHANGELOG.md): version history\n\n## License\n\nMIT. See [LICENSE](./LICENSE).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fkyaukyuai%2Falmanac","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fkyaukyuai%2Falmanac","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fkyaukyuai%2Falmanac/lists"}