bullswarm 0.25.5 → 0.27.0

package/AGENTS.md

@@ -20,37 +20,31 @@ content. Published as `bullswarm` on npm.
 5. Workflow dispatches must honor the same guarantees as single runs:
    `BULLSWARM_DEPTH` is propagated, burst-gated pools are excluded, and
    auth verdicts quarantine the pool + append to the shared decision log
-   (R6/R7/R8 in `src/workflow/runtime.js`).
-6. Adversarial verification is a first-class primitive: a `verify` step
-   reads a prior outFile and demands a JSON `{ok, concerns, summary}`
-   verdict before downstream steps can trust the work (R-skeptic).
-7. Workflows can be built incrementally from the shell
-   (`bullswarm workflow draft create/phase/step/set/...`). Drafts are
-   stored under `~/.bullswarm/drafts/<name>/` and are runnable by name
-   without an upfront JSON. JSON is still the durable artifact — drafts
-   are JSON documents, just built one mutation at a time.
-8. New goal workflows are caller-planned programs in a shared workspace.
+   (R6/R7/R8 in `src/workflow/v2-dispatch.js`).
+6. Adversarial verification is a first-class primitive: an action naming
+   requirements in `evidenceFor` is dispatched under an evidence contract and
+   judges them from the durable artifact, so a requirement is only verified by
+   work someone else inspected (R-skeptic).
+7. New goal workflows are caller-planned programs in a shared workspace.
    `bullswarm workflow goal --program` executes the graph; `--orchestrator`
    explicitly delegates planning. File territories are advisory scheduling
    hints, and the graph finishes without automatic gap rounds. `verified`
    separately records requirement evidence. `--isolation` opts into strict
    per-worker worktrees. Saved V2 runs preserve their original semantics.
+8. Historical authored-graph runs remain visible as read-only `legacy` rows.
+   Their executor was removed in 0.27.0; driving commands fail closed before
+   dispatch and historical run directories remain untouched.
 
 ## Development
 
 ```bash
 npm test            # full suite, no network needed (meters read from cache)
 node bin/bullswarm.js doctor --json   # readiness report
-node bin/bullswarm.js workflow list   # discover workflows
+node bin/bullswarm.js workflow goal "Fix the failing tests" --program plan.json
 node bin/bullswarm.js workflow runs   # ongoing workflow instances
 node bin/bullswarm.js workflow runs --all   # including historical
-node bin/bullswarm.js workflow validate <file>  # dry-run
-BULLSWARM_HOME=/tmp/bs node bin/bullswarm.js workflow run <file>   # sandboxed run
-# Build a workflow from the shell:
-bullswarm workflow draft create my-audit
-bullswarm workflow draft phase add my-audit discover
-bullswarm workflow draft step add my-audit discover list-files --type run --prompt 'List files'
-bullswarm workflow draft run my-audit
+# Validate a caller-authored program before launch:
+bullswarm workflow plan validate "Fix the failing tests" --program plan.json
 # Operate on a run by shortId (6 chars) or full runId (`wf-...`):
 bullswarm workflow runs show <shortId>
 bullswarm workflow runs delete <shortId> --yes
@@ -59,13 +53,13 @@ bullswarm workflow runs delete <shortId> --yes
 ## Using bullswarm from another agent
 
 If you are an agent that wants to offload bounded work via bullswarm,
-read `skill/SKILL.md` — that's the agent-facing user guide. Use
-`bullswarm delegate` (or the installed `/bullswarm` skill) by default: it
-previews whether one bounded agent or an autonomous workflow is appropriate,
-shows the conceptual plan, and executes the chosen engine. Reach for `run`,
-`workflow goal`, or a fixed workflow graph directly only when the caller has
-already chosen that execution shape. The skill is published alongside the
-package and is the canonical reference for the CLI surface.
+read `skill/SKILL.md` — that's the agent-facing user guide. There are
+exactly two ways to start work, and the caller chooses the shape itself: one
+bounded outcome goes to `bullswarm run`; parallel territories, integration,
+or independent acceptance go to `bullswarm workflow goal` with a program you
+author (`bullswarm workflow plan contract` returns the schema). There is no
+classifier or preview step. The skill is published alongside the package and
+is the canonical reference for the CLI surface.
 
 - Zero runtime dependencies. Node >= 18. Tests must never require network:
   prime `~/.bullswarm/meters/*.json` caches with fresh timestamps if needed.

package/CHANGELOG.md

@@ -1,5 +1,197 @@
 # bullswarm changelog
 
+## 0.27.0 — one workflow engine
+
+- Fixed: a worker that floods its stdout could kill the kernel. Every chunk of a
+  worker's stdout and stderr was appended to one string; a command-code worker
+  whose transcript outgrew Node's maximum string length threw
+  `RangeError: Invalid string length` inside the stream handler, the kernel died
+  as an uncaught exception, and every worker it supervised died with it (seen
+  twice on the same action in one day; the trace is in
+  `~/.bullswarm/goals/<runId>/stderr.log`). Streams are now captured through a
+  bounded buffer that keeps the first and last 16 MiB of each stream, counts what
+  it dropped (`captureTruncated` on the attempt observation), and any exception
+  raised while reading a worker now fails that attempt instead of the kernel.
+  Fatal-signature matching already looked only at the last 4,000 characters, so
+  quota and auth detection are unchanged.
+
+- There is now one workflow engine. The authored-graph verbs `workflow run`,
+  `validate`, `list`, `draft`, `inspect` and `approval` are gone — each falls to
+  the workflow-level unknown-verb message, exits 2 and spawns nothing — and with
+  them the eleven V1 modules they drove: `runtime.js` (1,902 lines),
+  `runner.js` (1,304), `draft-cli.js` (434), `validate.js` (366),
+  `decision.js` (357), `tui.js` (297), `draft.js` (268), `result.js` (239),
+  `template.js` (146), `schema.js` (81) and `semaphore.js` (56). The V1 panel
+  model, timeline and orchestrator-detail twins came out of `dashboard.js`
+  (−965), the V1 branches out of `cli.js` (−291), `runs-cli.js` (−141),
+  `short-id.js` (−127) and `watch-cli.js` (−132). `src/workflow/*.js` goes from
+  16,941 lines to 10,257. Also deleted: the five saved definitions
+  `workflows/adaptive-code-review.json`, `agent-model-comparison.json`,
+  `connector-audit.json`, `smoke-two-step.json` and `verify-and-cap.json`
+  (257 lines) together with the `workflows/` entry in package.json `files`;
+  `scripts/sanity-multi-claude.mjs` (131), which only exercised `runWorkflow`;
+  and two more tools that could only reach deleted modules —
+  `bin/check-output-schema.js` (30, the V1 `outputSchema` worker preflight) and
+  `scripts/planner-contract-probe.mjs` (175, a probe for the V1 `decide`
+  planner). `workflow capabilities` now reports `engines.authoredGraphs` as
+  `{ retired: '0.27.0', command: null }` instead of advertising a verb that
+  exits 2. `bullswarm workflow --help` describes one engine.
+
+- `newRunId` moved from `runner.js` into `src/workflow/short-id.js`, same
+  behaviour, exported; `v2-runtime.js` imports it from there.
+
+- Removed the last V1 remnants that no gate caught because they named no
+  deleted symbol: the dead authored-graph planner prompt in
+  `src/workflow/goal.js` (`AUTONOMOUS_ORCHESTRATOR_PROMPT`,
+  `PLANNER_RULES_SECTION`, `PLANNER_EXAMPLES_SECTION` — 36 lines whose only
+  consumer was the deleted `runtime.js`, and which still taught `type`,
+  `stepTemplate`, `itemsFrom`, `outputSchema`, `covers` and `completion.when`
+  to a planner the V2 validator would reject), and the permanently-zero
+  `fanout: { total, ok, failed }` counter that `dashboard.js` still put on every
+  row and rendered behind an unreachable branch. `outputSchema`, `itemsFrom` and
+  `stepTemplate` now appear nowhere in `src/`. The `skill/references/operations.md`
+  "Adversarial verification" section described the removed `{ok, concerns,
+  summary}` verify verdict; it now documents the `bullswarm.workflow.evidence.v2`
+  envelope the kernel actually enforces. `AGENTS.md` doctrine item 5 pointed at
+  the deleted `runtime.js` and now points at `v2-dispatch.js`.
+
+- Historical authored-graph runs stay readable, read-only, and nothing tries to
+  drive them. A run directory whose `state.json` lacks
+  `schemaVersion: 'bullswarm.workflow.state.v2'` is a legacy run: `workflow runs`
+  (with `--all` and `--json`) lists it as one row — short id, run id, name or
+  goal, status, age — marked `legacy`, reading only those five fields and never
+  throwing on a missing one, and the workflow home lists the same row. Every
+  driving command — `runs show`, `runs result`, `watch`, `cancel`, `resume`,
+  `steer`, `action show`, `tui <runId>` — prints exactly one line, `legacy
+  authored-graph run <shortId>: its executor was removed in 0.27.0; files remain
+  under <dir>`, and exits 2 before touching anything — including on the older
+  directories that hold only a `workflow.json` and never had a `state.json` at
+  all; the workflow home shows
+  that same line in its detail pane. `events <runId>` still replays the durable
+  JSONL and `runs delete <id> --yes` still removes the directory. Historical
+  directories are never modified. The stale-owner reconciliation that used to
+  run before every dispatch is gone with the V1 liveness model it served.
+
+- Tests: 818 -> 661. Six V1-only files were deleted
+  (`workflow-adaptive`, `workflow-gaps`, `workflow-draft`, `workflow-schema`,
+  `workflow-validate`, `workflow-run` — 151 tests); `workflow-runs`,
+  `workflow-watch`, `assignments`, `workflow-interruption`, `workflow-steering`
+  and `workflow-goal` were rewritten onto the V2 kernel keeping every assertion
+  about shared behaviour; `workflow-dashboard` went from 48 cases to 35 — 21
+  V1-only cases removed and 2 added with the rewrite, then 6 re-added against V2
+  fixtures for the rendering behaviours the removal had dropped (blocked-action
+  naming, one segment header per phase or dependency level, the mid-segment
+  continuation header, parallel levels grouped in declared order, the narrow
+  layout, and auto-follow); and a new `workflow-legacy-runs` (11 tests) proves
+  the legacy contract against a synthetic legacy `state.json`.
+  `tests/manual-dynamic-real.mjs` (330 lines), a
+  manual real-provider matrix for authored `run`/`decide` graphs, went with the
+  executor; it was never part of the suite count. No test dispatches a real
+  provider.
+
+## 0.26.0 — two entry points, kinds and rungs
+
+- There are now exactly two ways to start work, and `bullswarm delegate` is
+  gone. `delegate` existed to decide, on the caller's behalf, whether a request
+  needed one agent or a workflow — with `--mode auto` it spent a real
+  analyze-lane dispatch on a deterministic-then-LLM classifier before doing any
+  of the actual work. The calling agent already knows the shape of its own
+  request, so that round trip bought latency and quota, not accuracy. `bullswarm
+  run` is the single-agent entry point and `bullswarm workflow goal` is the
+  workflow entry point; `bullswarm --help` names those two, in that order, and
+  `bullswarm delegate` now exits 2 with the standard unknown-verb message and
+  dispatches nothing. The packaged skill and the awareness block registered into
+  Codex, Claude, and Grok say the same thing in six lines: decide the shape
+  yourself, there is no preview step.
+
+- `bullswarm run --lane analyze` now defaults to `medium` effort instead of
+  `high`. The V2 validator has always defaulted an analyze action to medium, so
+  the same lane meant a different tier depending on which entry point you used.
+  `run` and the validator now read one exported `DEFAULT_EFFORT_BY_LANE` table,
+  and `run --help` states the corrected default.
+
+- Program actions can now say what they *are* instead of restating how to route
+  them. The optional `kind` field takes one of seven values — `mechanical`,
+  `io-read`, `check`, `implement`, `integration`, `architecture`,
+  `adversarial-acceptance` — and derives `lane` and `effort` from a table
+  exported alongside the existing per-lane default. Resolution is per field: an
+  explicit action field wins, then the kind table, then a new optional
+  program-level `defaults` object (`effort` and `reasoning` only), then the lane
+  default. A kind outside the closed list is a validation error with the allowed
+  values named, because it is a typo in the program rather than a runtime
+  condition. Programs that use neither `kind` nor `defaults` and state
+  `lane` and `effort` on every action normalise byte-identically to before;
+  `effort` itself is now optional and falls back to the per-lane default
+  where it used to be rejected as missing.
+
+- Two non-blocking advisories, never rejections. `all-writers-high` fires when
+  three or more build/chore actions run and none is below high effort;
+  `docs-at-high` fires when a build/chore action owns only `*.md` files at high
+  effort. `workflow plan validate --json` carries them as `advisories` and the
+  human output prints `advisory:` lines; `workflow goal --program` prints the
+  same lines at launch. Neither changes acceptance or an exit code. The kernel
+  records them on the run, and `workflow runs show` lists them.
+
+- `workflow runs result`, `runs show`, and `action show` print `kind` next to
+  lane and effort when an action has one. `workflow action show` now understands
+  autonomous V2 runs at all — it previously read only the V1 action ledger and
+  failed on every V2 run.
+
+- **Rungs: one pool's model plus its reasoning level, for one effort tier, read
+  and written as one thing.** `bullswarm strategy rungs [--json] [--pool <name>]`
+  prints one row per enabled pool and configured tier with the effective model
+  and its source, the effective reasoning level and the layer that chose it, the
+  dated Epoch benchmark evidence for that model *at that level* (`blended`, cost
+  per task, tokens per task), and the local record from the decision log for that
+  pool and tier (dispatches, median wall minutes, ok share). Absent evidence
+  prints `no evidence` and an unmeasured tier prints `no dispatches`; neither is
+  ever estimated. `strategy inventory --json` gained the same rows under `rungs`.
+
+- `bullswarm strategy set-rung <pool> <tier> --model <model> [--reasoning <level>]
+  [--force]` writes both halves of a rung in one atomic state save, so the model
+  and the thinking depth can never land separately. A rung is singular per pool
+  and tier: the tier moves off whichever model held it, and that model keeps its
+  other tiers. A level the connector cannot express is clamped to the strongest
+  it accepts and the clamp is printed. An unknown pool or tier exits 2; a model
+  absent from the pool's cached discovery exits 2 and lists the known models
+  unless `--force` is given. Neither `rungs` nor `set-rung` ever spawns model
+  discovery. No state migration: rungs are a projection of `strategy.modelTiers`
+  and `strategy.reasoning`, and `~/.bullswarm/state.json` gained no keys.
+
+- The setup wizard's tier step now shows each suggested rung with its benchmark
+  evidence line and asks one reasoning question per configured tier. **Behavior
+  change:** Enter keeps that connector's own per-tier default and writes nothing,
+  where the previous question stored a suggested level (`xhigh`/`high`/`medium`)
+  on a blank answer. Connector defaults remain the final fallback, so a pool with
+  no configured rung behaves exactly as before. Non-TTY and `--yes` paths are
+  unchanged.
+
+- A dated evidence datapack per model *and reasoning level*, from Epoch AI.
+  `data/epoch-benchmarks.json` (schema `bullswarm.epoch.benchmarks.v1`) is built
+  by the new `scripts/refresh-epoch-benchmarks.mjs` from Epoch's cursorbench,
+  deepswe, arc-agi-2, and critpt exports, and `src/lib/epoch-benchmarks.js`
+  reads it with the same cache → bundled → URL fallback as the OpenRouter pack.
+  `rungEvidence()` returns the mean of whichever of those four scores exist for a
+  (model, level) pair as `blended`, with cost and tokens per task from
+  cursorbench; `normalizeModelId()` is how connector model ids match the export.
+  The data is used under CC BY 4.0 — Epoch AI, 'AI Benchmarking Hub'. Published
+  online at epoch.ai. Retrieved from https://epoch.ai/benchmarks.
+
+- The daily refresh job is renamed `.github/workflows/refresh-benchmarks.yml` and
+  now refreshes both assets on the rolling `benchmark-data-latest` release: it
+  downloads and unzips Epoch's public export, runs the script, runs the new
+  tests, and uploads `epoch-benchmarks.json` next to `openrouter-benchmarks.json`.
+  The ambiguous `npm run refresh:benchmarks` script is split into
+  `refresh:openrouter` and `refresh:epoch`, since only one of the two now
+  refreshes "the benchmarks".
+
+- The OpenRouter builder is unchanged, and that is a finding rather than an
+  omission: the Artificial Analysis records in the 2026-09-08 capture carry only
+  `agentic_index`, `coding_index`, and `intelligence_index` per model, with no
+  reasoning-effort marker on any of the `reasoning_effort`, `effort`, `variant`,
+  `reasoning`, or `reasoning_level` fields checked, so there is no per-effort row
+  to keep. `data/README.md` records the field names inspected.
+
 ## 0.25.5 — forecast-aware routing
 
 - Bullswarm now knows what it is already running. Every dispatch registers the