vinta-ai-workflows 0.7.0-alpha21 → 0.7.0-alpha23

package/CHANGELOG.md

@@ -65,7 +65,10 @@ the project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
     JSON on stdin and stdout). Out-of-tree adapters register through
     `registerSystemOneAdapter`. The file also turns on the built-in judges.
     This is the only API key the package uses; LLM harnesses still run on your
-    logged-in CLIs.
+    logged-in CLIs. Before a run starts, the preflight asks the classifier one
+    synthetic question, so a missing or rejected key, or an unreachable URL,
+    fails at minute zero rather than at the first gate. `"probe": false`
+    skips it. `doctor` warns about an `http` adapter with no `api_key_env`.
   - **Judge gates.** A workflow gate can be
     `{ "judge": { "question" | "question_ref", "labels", "fail_on",
     "threshold", "on_unavailable" } }` instead of `{ "cmd": … }`. It asks the
@@ -204,6 +207,47 @@ the project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ### Changed
 
+- **`plan-feature` AI model tiers pick up the late-September releases.**
+  `plan-feature/resources/ai-models.yaml` now cites `claude-sonnet-5-5`
+  (tier 2, was `claude-sonnet-5`) and `gpt-6.1-sol` (tier 3, was
+  `gpt-6-sol`). The retired `gemini-3-pro` (tiers 3–4) is replaced by
+  `gemini-3.1-pro-preview`, the only Gemini Pro id currently listed. Tier
+  *placement* is unchanged. **Consumers**: re-sync to pick up the refreshed
+  model suggestions.
+
+- **`vinta-ai-maestro run` starts the run as a background job and returns.**
+  The run no longer needs the terminal that started it, or any terminal: close
+  it and the run carries on. `run` prints the run id and the commands that
+  reach it, and if the preflight refuses it prints the report and says the run
+  did not start. **Migration:** a script or CI job that relied on `run`
+  blocking until the run ended, and on its exit code, adds `--foreground`,
+  which keeps the old behaviour exactly. The run's job serves no UI and prints
+  no URL any more — see `ui` below. (`vinta-ai-maestro`)
+- **New commands to reach a background run:** `status [run-id]` (every run, or
+  one run phase by phase; `--json` for scripts), `logs <run-id> [-f]` (the
+  run's job log — what `run` used to print — followed until the job exits),
+  `pause <run-id>` (nothing new starts, running phases finish their current
+  step, then the job exits with the run `paused`; `run --resume` continues it)
+  and `stop <run-id>` (live agent turns and gates are killed and the run ends
+  `cancelled`, which `--resume` refuses). Both also accept `--wait`. Two new
+  run statuses come with them, `paused` and `cancelled`; the post-mortem still
+  reports `done` or `failed`, and a paused run writes none until it ends.
+  (`vinta-ai-maestro`)
+- **`vinta-ai-maestro ui` serves the browser UI, and hosts no runs.** `serve`
+  is the same command under its old name. It reads every run from the journal
+  and forwards a live run's requests and socket to the job hosting it, so
+  watching, steering, answering and taking over work as before — and closing
+  it no longer interrupts anything. A run started or resumed from the UI is
+  launched as a background job with `ui`'s run settings. The job's own token is
+  kept in `.vinta-ai-maestro/runs/<run-id>/job.json` (mode `0600`, removed when
+  the job ends) and is never printed; treat that file like the URL.
+  (`vinta-ai-maestro`)
+- **`purge` keeps a run whose job is still running**, and says so.
+  (`vinta-ai-maestro`)
+- **The run view has Pause and Stop buttons**, beside Replay, while the run is
+  running. Stop asks for confirmation first, because it cannot be undone; both
+  say what is happening until the run's job has ended it. (`vinta-ai-maestro`)
+
 - **The canvas no longer swallows the page's scroll.** A bare wheel over the
   graph used to zoom it and block the page from scrolling; it now scrolls the
   page, and zoom is ⌘ or Ctrl + scroll (which is also how a trackpad pinch
@@ -213,6 +257,13 @@ the project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ### Fixed
 
+- **`check-ai-models` no longer mistakes sibling models for snapshots of a
+  cited id.** The id matcher used substring matching, so `claude-sonnet-5`
+  matched `claude-sonnet-5-5` (hiding the newer Sonnet and Opus releases).
+  It also matched `gemini-3-pro` to `gemini-3-pro-image`, so a retired model
+  still looked available. It now accepts only exact ids, dated or `-vN`
+  snapshot suffixes, and provider prefixes.
+
 - **The editor's canvas inspector no longer offers fields it then discards.**
   Its Status and Wave fields were folded back into the workflow and silently
   dropped — status is run state and wave is computed from dependencies — and

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vinta-ai-workflows",
-  "version": "0.7.0-alpha21",
+  "version": "0.7.0-alpha23",
   "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: 2026-09-16
+last_verified: 2026-10-02
 
 # Drift is checked against a free, no-key community aggregator (models.dev,
 # vendor-keyed with release dates + pricing), cross-checked against LiteLLM's
@@ -50,7 +50,7 @@ tiers:
       fixtures, simple HStore/ArrayField additions.
     models:
       anthropic:
-        - id: claude-sonnet-5
+        - id: claude-sonnet-5-5
           note: step-up target when phase touches >3 files
       openai:
         - id: gpt-6-luna
@@ -68,9 +68,9 @@ tiers:
       anthropic:
         - id: claude-opus-5-5
       openai:
-        - id: gpt-6-sol
+        - id: gpt-6.1-sol
       google:
-        - id: gemini-3-pro
+        - id: gemini-3.1-pro-preview
           note: stronger code recall
 
   - tier: 4
@@ -86,5 +86,5 @@ tiers:
         - id: gpt-6-astra
           note: frontier model; roughly twice sol's price, so reserve it for tier 4
       google:
-        - id: gemini-3-pro
+        - id: gemini-3.1-pro-preview
           note: with thinking