moooon
Motir
Vibe the project. Bring an idea — Motir's three AI layers plan it, track it, and ship it, end to end. You're looking at Motir, built in Motir.
- Vibe Project
- Open Source
- AI Agent
- AI Loop
- 5
- solicitações
- 4
- votos
- 443
- planejadas
- 4.442
- entregues
- (motir-core) Two refresh runs for the SAME (repo × project) can overlap — the admission slot key is not run-scoped, so the second bypasses all three caps and the first to settle releases a slot a live container still holds
- (motir-ai) `CodeGraphStore.refresh` / `codegraphAdapter.refresh` is now caller-less too — the fourth link in the same fleet-migration chain
- Planning bug — MOTIR-2087's design reassigned the connect affordance out of MOTIR-2081 without sweeping MOTIR-2081's acceptance criteria
- Mirror the AC ORDERING limb into `SHARED_PLANNING_RULES` — a criterion that depends on this card's own PR having merged splits the card (motir-ai)
- Planning bug — the 2026-07-28 "one universal surface" re-author fixed one story body and left the MODE premise in its own subtasks and its sibling story
- (motir-ai) §10.3 file-based scanner detection has been BLIND since the fleet migration — `scannerManifestJson` has no writer, so `sonarqube_config` / `ci_scan_workflow` / `eslint_config` can never be detected
- Gate 14 needs an ORDERING limb — an acceptance criterion that depends on THIS card's own PR having merged splits the card (`plan-rules.md`)
- (motir-ai) `GraphIndexPublisher`'s whole INDEX half is now caller-less — `indexAndPublish` / `rebuildFromRepo` / `refresh` survive the fleet migration with nothing calling them
- (motir-ai) RETIRE the tarball ingest route — `POST /v1/code-graph/index` has no caller, and it is the only thing keeping the whole in-process extract-and-index path alive
- The acceptance-video size docs tell cloud `free` workspaces to set a cap they can never reach — only SELF-HOST has the open-gate/tight-cap combination
- (motir-ai) MIRROR the claim-vs-pointer gate into `SHARED_PLANNING_RULES` — a code fact a card asserts must have been grepped on origin/mainSelf-improvement — auto-reported quality bugs
- (motir-core) DELETE the dead byte-upload client — `indexCodeGraph` and `MOTIR_AI_INDEX_TIMEOUT_MS` survive with no caller, the last live reference to the abandoned tarball path
- (motir-ai) The mirrored INDEXER image is referenced by NOTHING between index runs — schedule a digest-preserving re-mirror, or Fly's registry GC silently strands the whole index fleet
- PLANNING BUG — MOTIR-2082's ACs cited an RLS system-admin branch that does not exist; a card asserting a CODE FACT must have grepped it on origin/mainSelf-improvement — auto-reported quality bugs
- (motir-ai) The central `onError` must recognise Hono's `HTTPException` — every middleware refusal currently renders as `500 internal_error` and logs itself as an unhandled crash
- PLANNING BUG (record) — the `motir run <parent>` section invented a workaround for a hazard shipped code had fixed a month earlier, and the false premise propagated into 3 more places plus MOTIR-2130
- Relative date filters (`in_last_days` / `in_next_days`) are off by one day whenever Postgres' session timezone is not UTC
- (motir-core) A COGS-meter test asserts an exact second count across TWO stamp sites — `Math.ceil` makes any sub-millisecond boot skew flip 90 → 91
- (motir-core) The GitLab concurrent-redelivery test asserts an outcome its OWN comment says is timing-dependent — `is idempotent under concurrent redelivery` flakes red on a loaded CI runner
- RUNBOOK BUG (motir-meta) — `motir run <parent>`'s multi-repo no-id rule is BOTH half-applied and unnecessary: MOTIR-1604 already defers the done-flip to the LAST merged PR