bullswarm 0.27.0 → 0.27.1

package/CHANGELOG.md

@@ -1,5 +1,94 @@
 # bullswarm changelog
 
+## 0.27.1 — audit cleanup
+
+- Deleted the remaining dead symbols the 2026-09-09 audit listed as Tier A:
+  `recordAgentAction` and `classifyAgentProgress`, `aggregateUsage`,
+  `parseEvidenceOutput`, `fiveHourTier`, `REASONING_DEFAULT_TIERS`, the twelve
+  alias re-exports around `runAutonomousV2` /
+  `assertV2ResumeCompatible` / `validateEvidenceEnvelope`, and
+  `integrationBlock`. `'workflow-v1'` is no longer an assignment source.
+  `currentUsedPct` no longer reads a `weeklyUsedPct` field no producer writes,
+  and the identical `--json` ternary in `strategy-cli` collapsed to one
+  `JSON.stringify`. `fixtures/openrouter/` stays; a new
+  `tests/refresh-openrouter-benchmarks.test.js` runs the refresh script against
+  those fixtures instead of deleting them.
+
+- `state.json` is no longer last-writer-wins (D5). EVERY state writer goes
+  through one locked read-modify-write — take `state.lock`, reload FRESH,
+  mutate, atomically replace the file, release. That is `run`, `pools`,
+  `health` and the fixture migration plus the twenty-two remaining
+  load-mutate-save sites: fifteen in `strategy-cli` (set-rung, set-model,
+  reset-tier, set/reset-reasoning, configure, apply, auto off, assign,
+  clear-assignment, exclude/include-model, set-subscription and the persisted
+  refresh report), three in the strategy TUI, and four in `setup`
+  (`setup --yes`, both writing steps of the wizard, and the reasoning step).
+  `grep -n 'saveState(' src/` now finds only `src/lib/state.js`, which defines
+  it and calls it once, inside `updateState`. Waiters retry for 10 s and a lock
+  older than 30 s is taken over so a killed process cannot bench the file
+  forever. `run --dry-run` no longer refreshes strategy or hits the network
+  (D3). The fixture migration no longer force-disables a pool the operator
+  explicitly enabled (D1). Disabled pools are no longer polled (D6). When every
+  lane-capable pool was dropped by an empty tier allow-list, the routing reason
+  now says so instead of blaming missing capabilities (D7). `pools`, `health`
+  and `run --dry-run` leave `state.json` byte-for-byte alone when they have
+  nothing to change. `src/workflow/fsjson.js` — the re-export shim that carried
+  the workflow importers through the move of the atomic writer into
+  `src/lib/fsjson.js` — is deleted; both layers import `src/lib/fsjson.js`
+  directly.
+
+- The content gate no longer treats an error-shaped JSON object as an answer,
+  and long outputs are scanned at both the head and the tail for failure
+  patterns.
+
+- An unrecognized `--flag` is a usage error on every command: Bullswarm prints
+  `unknown flag --name` plus that command's synopsis and exits 2, before
+  self-initializing, routing, or spawning anything. `--lane` is required on
+  `run` (omitting it, or passing anything else, exits 2). `--limit` on
+  `workflow runs` must be a positive integer. `health --json` now selects the
+  machine-readable report; the default is a human summary of the same facts.
+  Help for `run --no-caller`, the always-JSON verbs, and the plan-contract /
+  plan-validate worker flags matches what the parsers actually accept.
+
+- A dead kernel is visible: `watch`, `runs show`, `runs result --json`, and the
+  TUI surface the last 20 lines of `stderr.log` instead of a silent stall.
+
+- Four small refactors from the audit's C1–C4 list: `TIER_LANES` is derived
+  from the 0.26 kind/effort tables so the strategy preview cannot disagree with
+  the validator; `clearTierAssignment` is the one writer of a cleared tier pin;
+  `finiteOrNull` is the one numeric coercion (blank/null prices and scores stay
+  unmeasured instead of becoming 0); quota signatures live in one table.
+
+- The OpenRouter datapack no longer pretends to have a bundled last-resort
+  file. Loaders try `~/.bullswarm/cache/` then the rolling GitHub Release; a
+  cache miss with no network yields an empty catalog. Epoch still ships
+  `data/epoch-benchmarks.json`.
+
+- Documentation made true: README no longer lists `runs cleanup`; bare
+  `bullswarm` is a TTY wizard and non-TTY callers self-initialize; the
+  OpenRouter/Epoch datapack fallback is described as it actually works; the
+  duplicated planning-targets paragraph appears once; `--name` is an exact
+  goal/name filter; `--classic` is V2-only and legacy watch exits 2;
+  `workflow-v1` is gone from the skill; `workflow goal` is no longer described
+  as an LLM-at-every-checkpoint loop; `delegate` is marked historical in the
+  2026-09-06 design note; GOAL.md and the ten dated `docs/` files that needed
+  a banner carry one.
+
+- Tests: 661 -> 713, 0 failures. Eight new files carry the new behaviour:
+  `unknown-flags` (11 — one bogus flag per parser, the two typed inputs, and a
+  drift guard that re-extracts every documented command form), `cli-run` (6 —
+  the D3/D1/D7 CLI contracts), `config` (5 — D6), `num` (4 — strict
+  `finiteOrNull`), `workflow-dead-kernel` (4), `state-race` (1 — an operator
+  write during a live run), `state-lock-sites` (3 — two real `strategy`
+  processes racing one home, four issued at once, and a `configure` document
+  that throws mid-mutation, writing nothing and freeing the lock) and
+  `refresh-openrouter-benchmarks` (1 — the refresh script against
+  `fixtures/openrouter/`). Existing files lost the
+  cases that only covered deleted symbols (`agent-events` 19 -> 13, and the
+  `fiveHourTier` case in `route`, which kept its
+  `FIVE_HOUR_NEAR_LIMIT_PCT === 75` assertion) and gained coverage for the
+  locked state writers, the fixture-migration rule, and the content gate.
+
 ## 0.27.0 — one workflow engine
 
 - Fixed: a worker that floods its stdout could kill the kernel. Every chunk of a

package/GOAL.md

@@ -1,6 +1,8 @@
 # bullswarm — Goal Statement
 
-**Status:** PROTOTYPE · **Owner:** cowcow02 · **Created:** 2026-08-21
+> Historical (2026-08-21): accurate when written; see CHANGELOG for what changed since.
+
+**Status:** HISTORICAL PROTOTYPE CHARTER · **Owner:** cowcow02 · **Created:** 2026-08-21
 
 ## One sentence
 

package/README.md

@@ -88,7 +88,7 @@ re-delegating and creating recursive swarms.
 ## Quick start
 
 ```bash
-bullswarm          # first run: interactive setup wizard
+bullswarm          # first run: interactive setup wizard on a TTY; non-TTY callers self-initialize
 bullswarm setup    # interactive provider/model configuration
 bullswarm setup --wizard  # broader worktree + integration questionnaire
 bullswarm pools    # meter state, pace position, quarantine status
@@ -98,7 +98,7 @@ bullswarm run --lane analyze --add-dir ~/some-repo --prompt "Inspect the parser"
 bullswarm workflow plan contract "Fix the failing tests and verify the change" --cwd ~/some-repo --json  # you are the planner
 bullswarm workflow goal "Fix the failing tests and verify the change" --cwd ~/some-repo --program plan.json
 bullswarm workflow goal "Fix the failing tests and verify the change" --cwd ~/some-repo --orchestrator auto  # dispatch a planner agent
-bullswarm health   # re-judge saved outputs; catch gate failures
+bullswarm health --json   # re-judge saved outputs; catch gate failures (omit --json for a human summary)
 ```
 
 ## Verbs
@@ -113,7 +113,7 @@ bullswarm health   # re-judge saved outputs; catch gate failures
 | `strategy` | Interactive provider/model control center with live high/medium/low route previews and an agent-facing JSON API |
 | `doctor` | Machine-readable readiness report; self-heals on first call |
 | `workflow` | Plan, execute, observe, and operate one autonomous workflow engine and its live instances. |
-| `runs` | Short alias for `workflow runs`, including list, show, result, delete, and cleanup operations. |
+| `runs` | Short alias for `workflow runs`, including list, show, result, and delete operations. |
 | `version` / `--version` | Print the installed Bullswarm version. |
 | `release` | Run the guarded local version-bump, commit, and tag workflow used before CI publishes to npm. |
 
@@ -203,9 +203,12 @@ Installed CLIs download only those public files and never need or receive an
 OpenRouter key.
 The sources are OpenRouter's [benchmarks API](https://openrouter.ai/docs/api/api-reference/benchmarks/list-benchmarks)
 and [models API](https://openrouter.ai/docs/api/api-reference/models/list-all-models-and-their-properties).
-The CLI caches the datapack under `~/.bullswarm/cache/`; network failure falls
-back to a stale or bundled datapack, then connector metadata, without blocking
-setup.
+The CLI caches each datapack under `~/.bullswarm/cache/`. OpenRouter is
+cache-or-network only: a fresh cache is used as-is, otherwise the rolling
+release is fetched, and a cache miss with no network yields an empty catalog
+plus connector metadata — there is no bundled `data/openrouter-benchmarks.json`.
+Epoch keeps `data/epoch-benchmarks.json` as a bundled last-resort, so a missing
+network never blocks setup when that file exists.
 
 The TUI lists every detected provider/account separately so its toggle matches
 its own quota meter. Enter drills into that provider's detected models. In the
@@ -343,7 +346,9 @@ caller asks for it by name.
 planning targets. They encourage the Workflow Planner to consolidate optional
 work, but the kernel never stops or rejects essential work merely because a
 target was reached. `--concurrency` still bounds simultaneous dispatches so
-the scheduler can batch a wider useful program safely.
+the scheduler can batch a wider useful program safely. There is no default
+wall-clock timeout: fresh semantic/transport heartbeats allow a useful worker
+to continue, while silence is inspected rather than blindly killed.
 
 The caller authors a complete program, or explicitly asks for a dispatched
 planner. The kernel validates the graph, executes it, and returns every action
@@ -502,12 +507,6 @@ per provider — the first as the primary `opencode2` pool, each additional one
 as its own `opencode2:<id>` pool — which is what the `--worker-model
 kaihk/gpt-5.6-luna` example above locks onto.
 
-`--max-agents`, `--max-actions`, and `--max-expansion-rounds` are soft V2
-planning targets: they guide the planner toward a small program but do not
-hard-stop useful work. `--concurrency` is the actual bound on simultaneous
-dependency-ready dispatches. There is no default wall-clock timeout: fresh
-semantic/transport heartbeats allow a useful worker to continue, while silence
-is inspected rather than blindly killed.
 New goal runs use the shared workspace regardless of the older setup
 worktree-isolation preference. Add `--isolation` to `workflow goal` when you
 explicitly want per-worker worktrees and strict ownership before integration.
@@ -569,7 +568,7 @@ no `0/1/i/l/o`). The full `wf-...` runId stays the durable handle.
 bullswarm workflow runs                    # ongoing only (default)
 bullswarm workflow runs --all              # ongoing + historical
 bullswarm workflow runs --historical       # only historical
-bullswarm workflow runs --name audit-code  # filter by workflow
+bullswarm workflow runs --name audit-code  # filter by exact goal/name
 bullswarm workflow runs --all --since 7d   # initiated in the last 7 days
 bullswarm workflow runs --historical --since yesterday --until today
 bullswarm workflow runs --all --from 2026-08-20 --to 2026-08-27

package/data/README.md

@@ -7,9 +7,14 @@ secrets and are never written here or shipped with the CLI.
 Both datapacks are published as replaceable assets on the
 `benchmark-data-latest` GitHub Release by
 `.github/workflows/refresh-benchmarks.yml`. Installed CLIs try
-`~/.bullswarm/cache/` first, then the copy bundled in this directory, then
-that release URL. A missing network never blocks setup when a bundled file
-exists.
+`~/.bullswarm/cache/` first, then that release URL.
+
+OpenRouter is cache-or-network only: there is no bundled
+`data/openrouter-benchmarks.json`. A cache miss with no network yields an
+empty catalog (and connector metadata), and never blocks setup.
+
+Epoch keeps `data/epoch-benchmarks.json` as a bundled last-resort, so a
+missing network never blocks setup when that file exists.
 
 ## `openrouter-benchmarks.json`