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
- 6
- requests
- 4
- upvotes
- 517
- planned
- 4,919
- shipped
- Apply the approved warm touches to the `motir` palette — light + dark tokens in `theme.css`, the separation / contrast suites over the new values, and `docs/palettes/motir.md` (motir-core)
- The default Motir palette gains a few warm-orange touches — Epic/Design chips, progress and tints warm up while the ink button, blue links/focus and ink logo stay
- Planning surface opens INSIDE a target that has no children — a childless story, task or bug lands on an empty level instead of beside the item
- A repo being re-audited says "started 2 weeks ago" — the in-progress row shows the PREVIOUS audit's time as when the new audit started
- motir-gateway's Fly image says it was built from `main`'s newest commit when it wasn't — on a `workflow_run` deploy, `GH_SHA` records the branch tip while the build checks out the commit CI tested
- Every card merged to motir-gateway fails CI on `main` and its deploy is skipped — `commit_lint` rejects the `MOTIR-<n> type(scope): …` squash-merge titles the card runs produce
- Two code-graph reads of the same cold graph at once: one waits and succeeds, the other fails `hydrate_failed` after 33 s — inside MOTIR-7619's 3-minute warming wait — and the convention job drops the cause from its error
- Motir inside Claude — a connector you sign into with your Motir account, one Claude Code plugin, and both listed in Claude's directories
- (motir-ai) graphRouting e2e "a recorded pointer's notice reaches the repository's two replicas only" sometimes gets 200 instead of 202 from a replica's refresh notice
- To fix is keyed per work item, not per run — one dead or red parent run shows up as a row on the parent and on every leg card
- The first read of a large code graph a graph-service machine doesn't hold fails — the download runs inside a 10-second request, so motir-core's audit says "not measured" until a retry finds the cache warmA dedicated code-graph service serves every tenant's graph reads, so any number of planning sessions read current code and the planning machines hold no graphs
- Adding a planning model asks for a typed model id — the dialog should offer every model the gateway serves that can plan (chat, planning-rated) and is not yet listed
- NotPlatformStaffError: The acting principal has no platform standing
- motir-ai `Coverage (planning pool)` fails at random: the branch threshold on `planJobService.appendProgress` depends on a real concurrent-append race actually happening
- (motir-ai · test) The planning-pool peer-wake seam test is flaky: machine B's start-up claim tick can land while queue(4) is inserting and take jobs meant for machine A
- (motir-core · design) To fix keyed by RUN — one Workbench entry per dead or red run with the cards it carries under it, one repair, a count of runs, and the per-card tag and banner pointing at that entry (en + zh) — deltas on `design/workbench/` and the To fix tag-and-banner design
- Design — the warm-orange touches in the default Motir palette: which `--color-*` / `--el-*` roles warm up, light + dark, measured against every hue they sit beside, drawn on the real board, chips and `--el-*` showcase (motir-core)
- Hosted planning is refused at the gateway for every audience — the seeded planner model `claude-opus-5-5` is on neither the planner token's allowlist nor the Anthropic channel
- (motir-ai) 15 of motir-ai#643's failures come from its MIGRATION, not its fixture — rate tests pin the rates in force for models the sync re-prices, and two pin decisions the sync now contradicts
- (motir-ai) The first schemaVersion-4 rate sync (#643) fails Vitest and five coverage jobs — rate tests still treat `tests/fixtures/upstream-prices.json`, which the sync rewrites, as a frozen 4-model v3 input