18 KiB
18 KiB
Mosaic Fleet — NORTH STAR
Generated file — do not edit by hand. Projected deterministically from
NORTH_STAR.yamlby the pure generator inpackages/mosaic/src/commands/fleet.ts(renderNorthStarMarkdown). Edit the YAML, then regenerate. Self-contained Mosaic.
Mission
A self-driving Mosaic system that 24/7 unattended converts a machine-readable goal set into merged, CI-green, budget-bounded change — looping plan→backlog→assign→execute→verify→merge→reassess — on Mosaic's OWN native backlog/dispatch engine. Mosaic is general-purpose: the user declares the system type they want (software delivery, personal assistant, research, business/operations, …) and the orchestrator provisions the matching persona roster and structure; the delivery fleet is one profile among many.
Substrate
The Mosaic Backlog is the backlog of record + dispatch engine, built on Mosaic's native Postgres storage service (@mosaicstack/db drizzle; PGlite-embedded by default, full Postgres by config).
Standing objectives
- NS-1 — Single machine-readable source (this file) drives planning; prose docs are projections.
- NS-2 — Every backlog item is an independently-shippable unit with stable id, priority, depends_on DAG, represented as a Mosaic Backlog card; spend tracked as advisory projection.
- NS-3 — The supervisor guarantees movement: no idle agent while ready dependency-satisfied work exists; no empty backlog without a replan request; assignment via Mosaic native dispatch/claim.
- NS-4 — Exactly one merge-gate approver; nothing reaches main except via pr-merge.sh after pr-ci-wait.sh success; Gitea branch protection is the backstop.
- NS-5 — Every unit bounded by wall-clock TTL on its claim; token caps enforced only where a real meter exists, else advisory.
- NS-6 — Context cleared between tasks for ephemeral runners (reset_between_tasks); persona+mission re-injected per task.
- NS-7 — Meta-loop (session-review + enhancer) continuously proposes small fleet-improvement PRs.
- NS-8 — Single operator-flippable PAUSE kill-switch (fleet/run/PAUSED) honored before every dispatch and every merge.
- NS-9 — Mosaic is a general-purpose multi-agent system: the user declares the SYSTEM TYPE to run (e.g. software delivery, personal assistant, research, business/operations) and the orchestrator provisions the matching persona roster and org structure from a cross-domain baseline persona library; the delivery/coding fleet is one profile among many.
- NS-10 — An adoption is not complete until the mechanism it replaces is removed. Two live conventions for one concern is the defect, not a transition state. Measured 2026-08-20: brain-home adopted by 9 modules and not 10; MOSAIC_HOME honored in 4 places, each re-deriving it; backlog cards declared while TASKS.md files stayed authoritative. Every one was decided correctly and left half-applied.
Success criteria
- AC-NS-0 (tier 0) — The operator launches an agent on any configured harness with one command, observes its state and sends it work without attaching to a terminal multiplexer.
- AC-NS-1 (tier 1) — The supervisor keeps a two-agent floor (1 orchestrator + >=1 enhancer) healthy across reboot.
- AC-NS-2 (tier 1) — A goal added to this YAML is decomposed to cards and either merged or escalated, with no human in the loop.
- AC-NS-3 (tier 1) — No PR merges with failure/error/no-status/timeout CI, and none bypass pr-merge.sh.
- AC-NS-4 (tier 1) — TTL is enforced on claims; token caps remain advisory until a real meter exists.
- AC-NS-5 (tier 1) — Flipping fleet/run/PAUSED halts dispatch and merges within one tick.
- AC-NS-6 (tier 2) — A user can declare a system type and the fleet provisions the matching persona roster + topology from the baseline library, with no code change.
- AC-NS-7 (tier 2) — A user-customized persona (edited or added via the orchestrator) survives mosaic update: baseline reseed never clobbers user overrides.
Workstreams
| id | title |
|---|---|
| A | Substrate — Mosaic Backlog on native Postgres storage service |
| B | Supervisor — movement guarantee, two-agent floor, dispatch/claim |
| C | Planner — goal decomposition into independently-shippable cards |
| D | Merge-gate — single approver, pr-merge.sh after CI wait |
| E | Meta-loop — session-review + enhancer improvement PRs |
| F | Safety-rails — TTL claims, advisory spend, PAUSE kill-switch |
| G | Kill-switch — operator PAUSE honored before dispatch and merge |
| H | Personas & system profiles — cross-domain library, system-type provisioning, update-surviving customization |
| I | Operator surface — launcher, fleet visibility, reliable steering (tier 0) |
| J | Web control plane — browser surface over the gateway (tier 1) |
| K | Clients — desktop and mobile over the same backend (tier 2) |
| L | Auth profiles — per-provider accounts, per-session selection (tier 2) |
Goals (backlog projection)
| id | title | tier | phase | priority | depends_on |
|---|---|---|---|---|---|
| A1 | Machine-readable NORTH_STAR.yaml + Markdown projection | 1 | 1 | must-have | — |
| A2 | Mosaic Backlog schema + storage-service card store (drizzle/PGlite) | 1 | 1 | must-have | A1 |
| A3a | Card lifecycle — create/claim/release with stable ids + depends_on DAG | 1 | 1 | must-have | A2 |
| A3b | TTL-bounded claim enforcement (wall-clock) on cards | 1 | 1 | must-have | A3a |
| A4 | Advisory spend projection per card (degrades to TTL, no real meter) | 1 | 1 | should-have | A3a |
| B1 | Supervisor tick — readiness scan, two-agent-floor health check | 1 | 2 | must-have | A3a |
| B2 | Native dispatch/claim — assign ready dependency-satisfied work | 1 | 2 | must-have | A3b, B1 |
| B3a | Planner decompose — goal added to YAML → cards | 1 | 2 | must-have | A2, B1 |
| B3b | Replan request on empty backlog; escalate on no-decompose | 1 | 2 | should-have | B3a |
| G1 | PAUSE kill-switch + merge-gate honored before dispatch and merge | 1 | 2 | must-have | B2 |
| H1 | Cross-domain baseline persona library (exec, marketing, ops, research, assistant + engineering roles) | 2 | 1 | must-have | A1 |
| H2 | System-type profiles — declarative mapping of system type to persona roster + topology | 2 | 2 | must-have | H1 |
| H3 | System-type provisioning — user declares type; orchestrator instantiates the matching roster + structure | 2 | 2 | must-have | H2 |
| H4 | Update-surviving persona customization — ad-hoc edits/additions persisted in a PRESERVE-protected override layer (baseline merged with overrides) | 2 | 2 | must-have | H1 |
| A5 | NORTH_STAR schema validator — every goal's workstream declared, every workstream has a goal, every depends_on id exists, every tier has a success criterion; runs in CI beside the Markdown regeneration check | 0 | 1 | must-have | A1 |
| I1 | One home resolver — a single function resolving MOSAIC_HOME with a sane default, adopted by every module. Today brain-home.ts is imported by 9 modules while 10 still use DEFAULT_MOSAIC_HOME, and MOSAIC_HOME is re-derived ad hoc in 4 places. NS-10 applies - finish the adoption and delete the second path | 0 | 1 | must-have | — |
| I2 | mosaic fleet ps sees the fleet that is actually running. Three measured blockers: the roster declares socket mosaic-fleet which does not exist, the 18 live sessions are on the default socket, and nothing writes a roster because seats are launched outside the CLI. Make the socket configurable and the roster written at launch, or make ps read tmux + seat dirs directly |
0 | 1 | must-have | I1 |
| I3 | Migrate fleet steering onto mosaic agent send --verify (exists, FLEET-OBS-005, spec FR-5) and retire tools/tmux/agent-send.sh, which forges the sender (D33) and returns an uninformative rc (D16, D34). FR-5 predates those defects by a month | 0 | 1 | must-have | I1 |
| I4 | mosaic fleet absorbs what launch-seat.sh does and launch-seat.sh is deprecated: compose the prompt file set, force the skill set, wire the style hook, fail closed on any unreadable input, export per-seat git identity, and register the seat in the roster. launch-seat.sh was a manual method; it is the reference implementation, not the destination | 0 | 1 | must-have | I1, I5 |
| I5 | Harness probe matrix — verify a working prompt-injection path for claude, codex, opencode and pi, and refuse any runtime whose path is unverified. AC-NS-0 clause 1 ("any configured harness") rests on this. The probe work in docs/plans/2026-08-19_launch-seat-multi-runtime.md (brain, untracked) transfers; its launch-seat.sh target does not | 0 | 1 | must-have | — |
| I6 | Finish the heartbeat responder (FLEET-OBS-002, the only Phase-2 task still in-progress; spec FR-2). Health must mean "answered a heartbeat", not "pane alive" — pane state measured two seats wrong on 2026-08-20 | 0 | 1 | must-have | — |
| I7 | Independent review and live-fleet dogfood of the Phase-2 verbs (FLEET-OBS-008), then land them (FLEET-OBS-009). Implementation is done and verification is not; "done" in a task file frozen five weeks is not evidence | 0 | 1 | must-have | I2, I3, I4, I6 |
| I8 | Neutralize misleading documentation — supersede headers on docs that state a stale status, a false blocker or a retired mechanism. Cheap, and it is tier 0 because a stale doc does not merely fail to help an agent, it actively misroutes one. Rebuilding the documentation is a separate and later job | 0 | 1 | must-have | — |
| I9 | Study t3code's agent-attach and multi-provider auth methods and record what transfers. Reference only — Mosaic implements its own within the stack, never adopts the code and never takes the dependency. Informs HOW I/J/K/L are built, not whether | 0 | 1 | should-have | — |
| J1 | Web control plane over the gateway — fleet visibility and steering in a browser, same data source as I2 | 1 | 3 | must-have | I2 |
| K1 | Desktop and mobile clients against the gateway, authenticated | 2 | 4 | must-have | J1 |
| L1 | Per-provider auth profiles with per-session selection | 2 | 4 | must-have | I4 |
Assumptions (vetoable)
- ASM-1 (vetoable) — The Mosaic Backlog on the native Postgres storage service is the backlog of record.
- ASM-2 (vetoable) — Claude gate roles have no native busy status, so readiness = pane-idle + heartbeat.
- ASM-3 (vetoable) — Two-agent floor = 1 orchestrator + >=1 enhancer.
- ASM-4 (vetoable) — Baseline personas ship in framework/fleet/roles/ (reseeded on update); user overrides live in a separate PRESERVE_PATHS-protected layer and win on merge.
Spend
- advisory: true
- No per-task token meter yet; budgets degrade to TTL. Spend is tracked only as an advisory projection alongside each card.