vinta-ai-workflows 0.7.0-alpha14 → 0.7.0-alpha16

package/CHANGELOG.md

@@ -7,6 +7,8 @@ the project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [0.7.0] — YYYY-MM-DD
 
+<!-- pre-release: 0.7.0-alpha16 on 2026-09-16 -->
+<!-- pre-release: 0.7.0-alpha15 on 2026-09-16 -->
 <!-- pre-release: 0.7.0-alpha14 on 2026-09-16 -->
 <!-- pre-release: 0.7.0-alpha13 on 2026-09-16 -->
 <!-- pre-release: 0.7.0-alpha12 on 2026-09-15 -->
@@ -24,6 +26,27 @@ the project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ### Added
 
+- **A run no longer dies with the terminal that started it.** `vinta-ai-maestro
+  serve` accepts `POST /api/runs`, so a run can be submitted to a daemon that
+  was already listening and belongs to that daemon rather than to the shell that
+  asked for it; the request answers with the run id as soon as the run is
+  registered, not when it finishes. Previously `run` was the only way to start a
+  run and it hosted its own daemon, so closing the window killed the scheduler
+  and every agent under it — and left the journal claiming `running` for ever,
+  with nothing able to tell that row apart from a run still in flight.
+
+- **`vinta-ai-maestro run --resume <run-id>` picks an interrupted run back up.**
+  Phases already `done` stay done and are not dispatched again, their lane
+  worktrees are adopted rather than re-provisioned — so whatever an agent had
+  written and not committed is still there — and a phase that was mid-turn when
+  the process died runs again from the top of its pipeline, because the turn it
+  was in belonged to a process that is gone. The plan comes from the run's
+  frozen snapshot, never from the document it was written from, so editing that
+  file in the meantime cannot switch plans under a run already under way. A new
+  `run_resumed` event records each hand-over; it is deliberately not a second
+  `run_started`, which would have moved the run's `started_at` to whenever it
+  was last picked up and made every duration short by the length of the outage.
+
 - **`vinta-ai-maestro gate <gate-id>` runs a declared gate from inside an agent
   turn.** A phase used to run its full gate suite five or six times a round —
   implementer, reviewer, fixer, reviewer again, then the authoritative `gate`
@@ -453,6 +476,24 @@ the project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ### Fixed
 
+- **Closing a terminal on a run now records the interruption.** `run` and
+  `serve` handle SIGHUP as well as SIGINT and SIGTERM — SIGHUP is what a
+  terminal sends when its window closes, and with no handler Node's default was
+  to die on the spot, running no teardown and writing no `run_ended`. Both
+  commands now mark their live runs interrupted before exiting and print the
+  `--resume` line for each, so a closed window leaves something recoverable
+  instead of a row nothing will ever move again.
+
+- **A member's reviews were counted as phases they took.** `node_crew` is
+  written for both seats — the reviewer's claim carries `role: 'reviewer'`, the
+  implementer's omits it — and the staffing rollup read neither, so every phase
+  a member reviewed was folded into the count of phases they implemented. On any
+  run with reviewers that inflated the number the rollup exists to put beside
+  the plan's estimate, and inflated `asPlanned`, which the run view divides by.
+  The two seats are counted apart now rather than one of them dropped: a review
+  is real work by a real member, and a member who only reviewed is neither a
+  phase-taker nor idle.
+
 - **An agent taking a lease left the resource panel showing the one before it.**
   The scheduler's own gate-pool transitions are journalled as `gate_pool`
   events, so the browser re-reads the snapshot when one lands. An agent's lease,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vinta-ai-workflows",
-  "version": "0.7.0-alpha14",
+  "version": "0.7.0-alpha16",
   "description": "Bootstrap AI tooling (AGENTS.md, skills, sub-agents) across Claude Code, Codex, Cursor, and VS Code + GitHub Copilot.",
   "type": "module",
   "bin": {

package/skills/vinta-derive-skills/resources/foundation-skills/plan-feature/resources/ai-models.yaml

@@ -11,7 +11,7 @@
 # Tier *placement* is human/LLM judgement — the job flags drift and proposes
 # edits, it does not blindly reassign tiers. See scripts/check-ai-models.mjs.
 schema_version: 1
-last_verified: 2026-07-13
+last_verified: 2026-09-16
 
 # Drift is checked against a free, no-key community aggregator (models.dev,
 # vendor-keyed with release dates + pricing), cross-checked against LiteLLM's
@@ -56,7 +56,7 @@ tiers:
       openai:
         - id: gpt-5-mini
       google:
-        - id: gemini-3.5-flash
+        - id: gemini-3.8-flash
 
   - tier: 3
     label: multi-file orchestration, business logic, SQL views
@@ -82,10 +82,12 @@ tiers:
       tuning slow query against partitioned hot table, debugging heisenbug.
     models:
       anthropic:
-        - id: claude-opus-4-8
+        - id: claude-opus-5
       openai:
+        - id: gpt-6-astra
+          note: frontier model; roughly twice sol's price, so reserve it for tier 4
         - id: gpt-5.6-sol
-          note: Frontier model for complex professional work
+          note: cheaper frontier option when astra's step up is not warranted
       google:
         - id: gemini-3-pro
           note: with thinking