Skip to content

CHANGELOG.md

Latest commit

 

History

History
243 lines (224 loc) · 15.7 KB

CHANGELOG.md

File metadata and controls

243 lines (224 loc) · 15.7 KB
Raw

Changelog

All notable changes to this project are documented here. The format is based on Keep a Changelog and this project adheres to Semantic Versioning.

[Unreleased]

Security

  • Raised the transitive overrides security floors to clear every open HIGH advisory in the locked tree: brace-expansion >=5.0.8 (GHSA-3jxr-9vmj-r5cp, GHSA-mh99-v99m-4gvg), fast-uri >=4.1.1 (GHSA-v2hh-gcrm-f6hx, GHSA-4c8g-83qw-93j6), js-yaml >=4.3.0 <5 (GHSA-52cp-r559-cp3m) and postcss

    =8.5.25 (GHSA-r28c-9q8g-f849). Also cleared the remaining moderates: hono >=4.12.32, protobufjs >=8.6.6, and @modelcontextprotocol/sdk 1.30.0 with @hono/node-server 2.0.12 (GHSA-frvp-7c67-39w9). npm audit now reports 0 vulnerabilities at the root and in board/, and the Trivy image gate passes with no fixable CRITICAL/HIGH.

Changed

  • ossf/scorecard-action 2.4.3 → 2.4.4.

Fixed

  • CI: the Upload OSV SARIF step no longer fails with "Path does not exist: osv.sarif" when an earlier gate fails — it is now guarded on the file existing, so a failing audit reports one real error instead of two.

[2.0.0] - in progress (feat/v2.0)

Major release — breaking changes are documented in MIGRATION.md. Authoritative plan: KAIBAN-v2.0-MASTER-PLAN.md.

Added

  • A2A v0.3 federation via @a2a-js/sdk: the gateway answers message/send, message/stream (SSE), tasks/get, tasks/cancel on POST /a2a/rpc; AgentCard v0.3 (object capabilities + abilities in skills[]). (BETA.1, ADR-015)
  • MCP server — first-party Model Context Protocol surface (Tools / Resources / Prompts / Elicitation) over Streamable HTTP. (BETA.2, ADR-017)
  • Resilience — the single-active orchestrator is promoted to src/shared (reusable, published): Redis checkpoint/resume, liveness/readiness probes, graceful drain, DLQ replay. (BETA.2, ADR-018)
  • Economics / FinOps — fleet-wide rate + cost control (token bucket + cost reservation), default-off. (BETA.3, ADR-019)
  • Governance Action Gate — default-off, non-bypassable when enabled: hash-chained audit, policy-as-code, kill-switch; hot-path enforcement wired into AgentActor. (BETA.3, ADR-020/021)
  • Universal AMQP driver seam — declared amqplib seam, unimplemented stub (coverage-excluded). (BETA.1, ADR-016)
  • dispatchToAgent actor-mailbox primitive in src/shared (replaces the removed tasks.create RPC). (The examples also gain a local parseAgentCardSkills helper to read v0.3 AgentCards — example code, not a published library export.)
  • Workflow budget guard (MAX_WORKFLOW_COST_USD / MAX_WORKFLOW_TOKENS) in both example orchestrators — checked between phases and before each revision; graceful STOPPED on breach (default 0.50 in the example compose files; 0 = unlimited; separate from per-agent MAX_TOKEN_BUDGET).
  • Playwright visual baselines for the React board + the two static example viewers (cd board && npm run test:visual).
  • scripts/smoke-consumer.sh — packs the Apache tarball and verifies a fresh consumer can import the public surface from both entry points (. and ./shared).
  • Packaging: ./shared subpath export + two-entry api-extractor; staging-dir npm pack (Apache-only artifact, no GPL leak). (BETA.1, ADR-011)
  • COMPLIANCE cross-walk and the v2.1 roadmap.

Changed

  • License (BREAKING): the published npm library is now Apache-2.0 (was GPL-3.0); the full application / board / examples remain GPL-3.0 (dual-license — see LICENSING.md, ADR-011).
  • Dependencies: KaibanJS 0.24.2, TypeScript 6.0, OpenTelemetry 0.219/0.77, bullmq 5.79, dotenv 17 — all latest stable, 0 vulnerabilities (ADR-012).
  • gateway / worker ROLE split — a single image runs as ROLE=gateway|worker. (BETA.1, ADR-013)
  • AbortSignal cancellation — an in-flight LLM call is aborted on task timeout / tasks/cancel (the bridge owns the LLM instance). (BETA.1, ADR-014)
  • Examples migrated to A2A v0.3 — removed all tasks.create / tasks.get / agent.status usage; both examples dispatch via dispatchToAgent and read AgentCard skills[].
  • Gateway HITL delivery — the durable per-task BRPOP-list write now precedes the pub/sub publish, and the board is ACK'd only after both succeed (a missed pub/sub message stays recoverable via the list fallback).
  • CompletionRouter subscribes lazily (on the first wait()) — a router that never waits no longer consumes the shared completed queue (fixes a competing-consumer hang between the gateway's A2A-executor router and an orchestrator router).
  • Kafka driver now throws a clear error on a 2nd subscribe() (explicit one-topic-per-driver contract) instead of silently breaking.
  • BullMQ driver sets job-retention defaults (removeOnComplete / removeOnFail, bounding Redis growth) and registers a worker error listener.
  • A2A executor logs the underlying error server-side on failure (the wire response stays generic).

Fixed

  • HITL re-arm loop — the terminal prompt no longer re-arms after a decision arrives (board OR terminal) or after stdin EOF; fixes the REVISE infinite re-prompt spin (100% CPU, process never exits, Ctrl-C ineffective) and restores the second HITL gate on the revised draft.
  • Board hangs on RUNNING after a hard failure — both orchestrators now publish a terminal STOPPED state on error, so the board reflects the failure instead of hanging.
  • Board store — malformed state deltas with no agentId / taskId are skipped (the Zustand map is never keyed by "undefined").
  • Fan-out/fan-in result↔index mismatch (global-research, plan finding C1/HIGH) — search results are now mapped to their dispatch index by taskId (was indexed by waitAll completion-order position), so searchers that finish out of order get the correct sub-topic, node label and logged taskId.
  • Governance Action Gate fails closed on a throwing validator; MCP-without-auth warning.
  • Byte-accurate data caps — the 64 KB outbound-message and 20 KB state-event-result caps are now measured in UTF-8 bytes (Buffer.byteLength), truncating on a codepoint boundary (was UTF-16 .length, which let multi-byte payloads exceed the byte cap).
  • Structured agent output — the KaibanJS bridge now JSON-stringifies a non-string (object) LLM result instead of emitting "[object Object]".
  • Model-pricing accuracy — estimateCost normalizes OpenRouter slugs / dated suffixes (openai/gpt-4o-mini, gpt-4o-2024-08-06, anthropic/claude-3-5-sonnet) before pricing lookup, and warns on a default-pricing fallback (was exact-match only → slugs silently mis-priced).
  • Config robustness — numeric env vars are parsed NaN-safe (AGENT_TIMEOUT_MS=abc no longer yields setTimeout(…, NaN) / instant timeouts).
  • Actor robustness — a timed-out task handler's late rejection no longer surfaces as an unhandledRejection.
  • Global-research budget guard now runs after every phase (search, write, governance, editorial), matching blog-team.
  • A2A input validation hardened; de-stubbed IDLE / TODO placeholders. (BETA.1)

Security

  • The published library ships 0 HIGH/CRITICAL advisories; release flow keeps SBOM + SLSA provenance + cosign signing (ADR-012).

[1.5.0-beta] - 2026-06-16

Pre-merge hardening pass (audit follow-up)

  • Container image CVEs (cleared the new Trivy gate to 0 fixable CRITICAL/HIGH): Dockerfile runs apk upgrade --no-cache in both stages (self-heals base-image openssl/musl/zlib — incl. a CRITICAL openssl); installs with npm ci (lock-faithful, no transitive drift above the audited 0-high lock) and pins glob >= 10.5.0 via overrides; and removes the bundled npm CLI from the runtime stage (the node base image's vendored npm shipped HIGH-CVE glob/minimatch/tar that the runtime — which only runs node — never uses). Also shrinks the image and reduces attack surface.
  • Kafka poison-message resilience: the consumer now skips an unparseable (non-JSON) record (logs a warning, advances the offset) instead of throwing and crash-looping on it — matching the BullMQ path's tolerance.
  • Trace-header parity: extracted a shared sanitizeTraceHeaders (drops non-string entries + rejects malformed traceparent); both BullMQ and Kafka drivers now use it (previously only BullMQ validated).
  • Timer leaks fixed: AgentActor clears its per-task timeout when the handler wins the race; SocketGateway clears the token-expiry disconnect timer when a client disconnects early.
  • CompletionRouter: rejects a duplicate wait() for an in-flight taskId instead of silently overwriting (which hung the first waiter).
  • Token budget wired: startAgentNode now forwards MAX_TOKEN_BUDGET to the AgentStatePublisher (was loaded into config but never reached the worker path; 0 = unlimited, so default behavior is unchanged).
  • CI: added a Trivy image scan to the docker job (gate on fixable CRITICAL/HIGH, SARIF to the Security tab); engines.node >= 22; PR trigger now also covers develop; board job renamed to reflect it typechecks+builds (it has no eslint).
  • Docs: corrected a broken ASVS→SECURITY.md link, stale test counts in BLOG_TEAM_TEST/EXAMPLES, the --chaos/--searchers example (global-research, not blog-team), createKaibanTaskHandler/KaibanTeamBridge signatures in SPEC, security impl paths + JIT_TOKENS_ENABLED in ACTOR_MODEL, and the documented overrides list in SECURITY.md.

Security

  • Cleared all critical/high dependency advisories (was 1 critical + 8 high): bumped @opentelemetry/{sdk-node,auto-instrumentations-node,exporter-trace-otlp-http} to the patched 0.217.x line and extended package.json overrides (protobufjs, axios, @grpc/grpc-js, langsmith, ws, hono, qs, uuid, fast-uri, fast-xml-parser, ip-address). Remaining 2 advisories are moderate, dev/build-only (see SECURITY.md).
  • Added SECURITY.md (private reporting, key handling, supply-chain posture).

Added

  • CI/CD: ci.yml (lint, typecheck, enforced 100% coverage, real madge gate, board tests, npm audit blocking on HIGH+, CycloneDX SBOM, gitleaks, OSV, CodeQL, Docker build), nightly.yml (Redis/Kafka/security e2e), release.yml (SBOM + SLSA provenance + Sigstore cosign signing), scorecard.yml.
  • Governance: CONTRIBUTING.md, CODE_OF_CONDUCT.md, CHANGELOG.md, .github/dependabot.yml (root + board + actions).
  • @opentelemetry/api added as a direct dependency (was imported but undeclared).
  • Unit test for the bootstrap/composition root (src/main/index.ts).

Fixed

  • global-research: editorial score always rendered N/A because the phase parsed the governance reviewer's Compliance Score label instead of the editor's Accuracy Score — corrected, with a regression test.
  • Coverage gate was a no-op (thresholds mis-keyed for vitest 4 and coverage.all unset); now coverage.all + include: src/** + coverage.thresholds enforce a genuine 100% of src/**.
  • Architecture gate was a no-op (madge --circular src/ scanned 0 files); now uses --extensions ts (scans the TypeScript graph).

Changed

  • Reconciled open Dependabot PRs #17/#18 into a single controlled refresh.
  • Packaging: added src/index.ts public-API barrel; package.json main/types/exports/files for a clean library publish (ships dist/src only — examples/board/tests excluded); tsconfig.build emits declarations.
  • Removed 4 divergent duplicate vitest.*.config.ts (kept .mts + wired test:e2e:live).
  • Docs softened to defensible/accurate wording: CI/Scorecard-backed badges; "one of the first" (not "the very first in the world"); compliance → capability language with a "library, not a certified product" disclaimer; corrected backoff (linear), result cap (20 000), and circuit-breaker threshold (10).
  • release.yml now also builds, pushes, and cosign-signs the container image.

Security (additional)

  • Applied sanitizeDelta PII stripping on the worker AgentStatePublisher path (defense-in-depth; previously only the middleware path).
  • Added LICENSE-EXCEPTIONS.md (GPL-3.0 SaaS-vs-distribution + commercial licensing).

Observability & quality (additional)

  • OTel metrics: kaiban.message.processed counter + kaiban.message.latency histogram via a PeriodicExportingMetricReader (OTLP when configured, Console in dev); emitted from the actor's success/DLQ paths.
  • Anomaly events: recordAnomalyEvent wired at the actor level (circuit_breaker.rejected, firewall.blocked).
  • Benchmarks: bench/throughput.bench.ts + npm run bench (vitest bench) + nightly job; property tests (fast-check) for channel-signing.
  • Lint: --max-warnings 0; examples/ now linted (browser-globals override for the viewer JS) — npm run lint covers src + tests + examples.
  • Deploy docs/manifests: HELM gateway secretRef + accurate image-tag default; AWS ECS worker CMD note; Railway per-service guidance (no unsupported railway.json).
  • Structured logging (pino): src/shared/structured-logger.ts — JSON logs with PII redaction, child loggers, level via LOG_LEVEL, and opt-in LOG_PRETTY=true (pino-pretty) for readable local/demo output. Rolled out across all production modules (AgentActor, SocketGateway, GatewayApp, SlidingWindowBreaker, AgentStatePublisher, DistributedStateMiddleware, MCP client, KaibanAgentBridge, Telemetry) and the worker bootstrap (main/index.ts) + shared orchestration (agent-node, orchestrator-state-publisher) — all operational logging is now structured. The demo createLogger and HITL interactive prompts stay intentionally human-readable (terminal UX, not operational logs).
  • API-surface gate: @microsoft/api-extractor (api:check/api:update) with a committed etc/kaiban-distributed.api.md report; CI fails on undocumented public-API drift — completing the 6-gate verification protocol.
  • Chaos / broker fault-injection e2e: test:e2e:chaos pauses the Redis broker mid-flight and asserts every buffered publish flushes on recovery with zero dropped agent messages; fully isolated (own config, afterAll/finally always restore the broker) and wired into nightly.yml.
  • Mutation testing (Stryker): test:mutation over the pure src/domain/** layer — 96.61% mutation score (break gate 85%), nightly job + HTML report. Remaining survivors are provably-equivalent mutants. (Stryker introduced no new HIGH/critical advisories.)
  • Accessibility: board axe-core a11y tests (vitest-axe) across 6 components — 0 violations.
  • Container image scanning: CI now scans the built Docker image with Trivy (gates on fixable CRITICAL/HIGH OS+library CVEs, ignore-unfixed, SARIF to the Security tab) — closing the image-layer gap the lockfile scanners can't see.
  • Doc accuracy (final reconciliation): corrected the remaining stale claims — unit/e2e test counts (769 unit / 77 files, 65 BullMQ-e2e) across README/EXAMPLES/ BLOG_TEAM_TEST; firewall verdict shape ({ allowed, reason }); Kafka auto-create dev-vs-prod note; blog-team phase wait times (120/240/300 s); ADR-003 cost wording (estimated via MODEL_PRICING); "linear backoff" in ACTOR_MODEL.md. Declared ts-node as a devDep (the documented npx ts-node dev-run path). The in-repo audit matrix now shows zero open items.
  • Deployment: verified + hardened Railway/Vercel/AWS/Azure — removed Vercel phantom steps; gateway SOCKET_CORS_ORIGINS everywhere; worker LLM keys/commands; health-check scoping (gateway-only); WEBSITES_PORT for Azure App Service.