Skip to content

Releases: Agentplat/agentplat

AgentPlat 0.3.0-beta.8 — coordinated 65-package preview

Choose a tag to compare

@douglas-grishen douglas-grishen released this 14 Sep 02:13
0e1e34e

All 65 AgentPlat packages are published at 0.3.0-beta.8 on npm next. No latest tag was changed.

This release integrates the pending Morphogenesis coverage work, corrected diagnostic convergence projection and dependency/toolchain updates through PR #170. Published source: 0e1e34e8755bfdecf708314130cc2385b4797f95.

All 65 staged tarballs were downloaded and byte-compared against the originating GitHub artifact before individual npm 2FA approval. Preparation and OIDC staging succeeded. Manifest digest: sha256-Q2fJLx0/55orxToaPFgKXNfRUQd873q5RiiGyWvf5Vw=.

Approved-release verification passed registry bytes, signatures, provenance, tags, complete distribution, pnpm portable, pnpm PostgreSQL and npm Node 22.22.0 consumers. The consumer cohort covers 65 packages and 216 export subpaths.

The temporary Beta 8 npm self-review exception was removed after publication; independent environment review and no administrator bypass were restored. The permanent owner PR-review exception remains, with required CI intact.

Distributed Morphogenesis staging, prolonged host-loss qualification, the full registered empirical campaign and external downstream promotion validation remain separate obligations. Publication and bounded consumer checks do not establish production readiness.

Distribution record · Development reconciliation and deferred work.

AgentPlat 0.3.0-beta.7 — coordinated 65-package preview

Choose a tag to compare

@douglas-grishen douglas-grishen released this 11 Sep 21:33
b7a2a56

All 65 public AgentPlat packages are available at 0.3.0-beta.7 under the coordinated npm next tag. latest was not promoted.

Published source: b7a2a565644550a3b1a1b927eafa664595c219e8. This release includes governed A2A 1.0 interoperability and Agent Registry alongside the existing public cohort.

Registry integrity, signatures, provenance and tags passed verification. The original CI consumer step failed because its list omitted three packages; the corrected harness subsequently verified all 65 packages and 216 export subpaths with three independent clean local consumers. This announcement does not label the original workflow as passing or claim production readiness.

Detailed distribution and consumer evidence.

AgentPlat 0.3.0-beta.5 — durable Agent Room operations

Choose a tag to compare

@douglas-grishen douglas-grishen released this 22 Aug 12:18
dc16b22

AgentPlat 0.3.0-beta.5 completes the operational evolution of Agent Room while preserving PostgreSQL and Agent Room domain state as authority.

Highlights:

  • transactional Room message intake into durable coordination
  • continuous leased coordination worker with retries, heartbeat renewal, metrics and graceful drain
  • optional Temporal adapter with canonical activity binding, restart and continueAsNew
  • typed AgentPlat Handoffs from agent tool output
  • progressive dependency-aware Planner reconciliation
  • automatic Room-scoped LiveView and SSE discovery
  • PostgreSQL V2–V11 execution, registry, Handoff, coordination, contribution, knowledge, plan, membership, operational-stream and checkpoint migrations
  • optional Asana human-contribution projection adapter

Publication evidence:

  • 56 coordinated packages published at 0.3.0-beta.5 under npm next
  • 190 public export subpaths verified from clean pnpm and npm consumers
  • PostgreSQL registry consumer passed 14 conformance cases
  • real PostgreSQL V1–V11, multiprocess restart recovery and real Temporal Server scenarios passed before publication
  • public terminology, type, test, compatibility, release and pack gates passed

Documentation:

  • docs/agent-rooms.md
  • docs/agent-rooms-postgres-migration.md
  • docs/adr/0043-agent-room-operational-coordination-boundaries.md
  • docs/reviews/agent-rooms-final-gap-closure-2026-08-21.md

Evidence boundary: these checks establish source behavior, packaging and deterministic integration. They do not claim production availability, third-party service-level behavior or live Asana validation. Asana remains an optional reference adapter.

Full changelog: https://github.com/Agentplat/agentplat/blob/v0.3.0-beta.5/CHANGELOG.md

AgentPlat 0.3.0-beta.4

Choose a tag to compare

@douglas-grishen douglas-grishen released this 19 Aug 02:48
9a2fe7e

AgentPlat 0.3.0-beta.4 client release

Install

npm install @agentplat/collective-runtime@0.3.0-beta.4

The prerelease is published under the next distribution tag. Pin the exact
version in production-like evaluation environments.

New public surfaces

  • @agentplat/collective-runtime/governed-collective-runtime
  • @agentplat/collective-runtime/durable-runtime-state
  • @agentplat/collective-runtime/partition-operation
  • @agentplat/collective-runtime/degraded-effect-budget
  • @agentplat/collective-runtime/partition-reconciliation
  • @agentplat/collective-runtime/compromise-lifecycle
  • @agentplat/collective-runtime/compromise-authority-lifecycle
  • @agentplat/collective-runtime/forensic-preservation
  • @agentplat/collective-runtime/team-topology-transformation
  • @agentplat/collective-runtime/team-formation-strategies
  • @agentplat/collective-membership/coordinator-election
  • @agentplat/collective-control/mandate-continuity
  • @agentplat/collective-planning/mission-continuity
  • @agentplat/trust/evidence-fusion-strategy

Migration

Existing runtime constructors and fail-closed behavior remain unchanged. The
governed facade is opt-in. Use the reference-integrated profile when every
critical gate must be present, and configure durableStore plus
idempotencyLedger for restart-safe operation. Durable deployments should use
the asynchronous lifecycle methods: pauseAsync, resumeAsync,
safeStopAsync, recoverAsync, and abandonAsync.

Compatibility

Node.js 20.19.3 or newer is required. The official release workflow verifies
clean registry consumers under Node.js 20 with pnpm and Node.js 22 with npm.

Empirical study preregistration V5 — 9ed8b0a

Choose a tag to compare

@douglas-grishen douglas-grishen released this 11 Aug 02:13
9ed8b0a

Public, signed preregistration for the complete registered local collective-control study bound to commit 9ed8b0a. Campaign paper-study-v5. Registration digest sha256:f5561d880d7d0536e39454a17ef232bdd9f00735c4940bdc7f025f68b6ec33aa. Attestation digest sha256:27a8c4ce6a2bd27c4ff8b351138d27cf97b39e7976bdb88053c56e4524833f28. Registered scale: 48 shards, 240 cells, 960 slots. External-spend ceiling: USD 0. Authorization not issued. Execution permitted: false. Results not collected. No shard was executed, no cloud resource was deployed, and this release grants no execution authority or empirical-result claim.

Collective capability attestation V1 — 9ed8b0a

Choose a tag to compare

@douglas-grishen douglas-grishen released this 11 Aug 02:12
9ed8b0a

Signed source-development evidence for commit 9ed8b0a. Covers the frozen V1 baseline of 19 capabilities. Source-development closure only; no empirical validation, execution authorization, or deployment certification.

Empirical study preregistration V4 — eb1ed4b

Choose a tag to compare

@douglas-grishen douglas-grishen released this 10 Aug 23:19
eb1ed4b

Public, signed preregistration for the registered local collective-control study bound to commit eb1ed4b. Campaign paper-study-v4. Registration digest sha256:2469fecd922781e5c54e805df229c85579922f8a2d6e1e57629b153cf5ee9399. Attestation digest sha256:d310a4d21ec85cf977a6584a2515cc83139d1c704a1d348485760607bcd01f89. Registered scale: 48 shards, 240 cells, 960 slots. External-spend ceiling: USD 0. Authorization not issued. Execution permitted: false. Results not collected. No shard was executed, no cloud resource was deployed, and this release grants no execution authority or empirical-result claim.

Empirical study preregistration V3 — 930dbcd

Choose a tag to compare

@douglas-grishen douglas-grishen released this 10 Aug 18:45
930dbcd

Public, signed preregistration for the registered local collective-control study bound to commit 930dbcd and its source-capability attestation.\n\n- Campaign: paper-study-v3\n- Registration digest: sha256:f6883e0fb6c0d846bb484ea2b60b63ff983c4658f277a0933ab55c86aa912d32\n- Attestation digest: sha256:817c62363d176333c1c889bb1044d261b83fb08b43dd363df56603dbcaadfa63\n- Registered scale: 48 shards, 240 cells, 960 slots\n- External-spend ceiling: USD 0\n- Authorization: not issued\n- Execution permitted: false\n- Results: not collected\n\nNo campaign shard was run, no cloud resource was deployed, and this release grants no execution authority or empirical-result claim.

Empirical study preregistration V2

Choose a tag to compare

@douglas-grishen douglas-grishen released this 10 Aug 13:28
ab803b3

Scope

Public, signed preregistration of Agentplat's finite-range collective-control study. This release fixes the protocol, source revision, normative adapter, evaluator, execution design, expected manifest, statistical policy, and zero-external-spend envelope before any observations are collected.

Immutable bindings

  • Source commit: ab803b3997751a2815e924c33fbe406086a20417
  • Registration digest: sha256:58c21b2f936ae59ff11ecf1d5ebe52943ca9cf797105070e532acfb9921ff78b
  • Attestation digest: sha256:68fd8734217e73129a2f64ec86cb01f6fd437a1ca73d8e4df22dc45ee7fa91a7
  • Adapter descriptor: sha256:5ff83f113c7ca5939788db542c049b3fd92a69b5d38b92b9ed018338e79c147d
  • Adapter implementation: sha256:2b3ba10b8f8951cd6d8573034a11d43e47ef540048be47e493a6b80487735bc4
  • Evaluator: sha256:9a2695bb843cb8d4db8d5aa168593a2d5b8a6c5c541f4b5dc974c9e441300439
  • Execution design: sha256:64190c54ebca39ec3b3b0c899b889c532a54c38b3ab64cf009a6fe3869f71801
  • Expected manifest: sha256:5a6ca359c6c562ec6eca11ac1dd29e27c3b198aa4762c63bfc4d123895c39f0c
  • Operation plan: sha256:88bde02898e97cb751f3c0173e4a3ef23869bb438087adacd1a538ef20baf0ef
  • Signing key: arn:aws:kms:us-east-1:184644493945:key/98c4767e-cc4f-4dd3-a32b-d3e2b5658181

Registered scale and budget

  • 48 independently authorized shards
  • 240 registered cells
  • 960 expected execution slots
  • 3,296,000 maximum registered interactions
  • Aggregation seed: 20260810
  • Maximum external spend: USD 0
  • Cloud compute, managed databases, paid model calls, and paid data egress are not permitted

Status at publication

  • Authorization: not_issued
  • Execution permitted: false
  • Results: not_collected
  • Empirical result claims permitted: false

No study shard was executed, no cloud resources were deployed, and no npm package was published as part of this release.

Empirical study preregistration V1

Choose a tag to compare

@douglas-grishen douglas-grishen released this 10 Aug 10:37
b603269

Pre-results, machine-verifiable preregistration for the Agentplat collective-control empirical study.

Source commit: b603269
Scientific registration digest: sha256:5682f7c8e0d7fd11019e77188b183811235443b1b434934e3fbd88681ec2638c
Attestation digest: sha256:4914beb3130a5b4e407a5b882e2e4ba40648e39c66f7856afbe98b0f5f239e8e
Signer: AWS KMS Ed25519 key arn:aws:kms:us-east-1:184644493945:key/98c4767e-cc4f-4dd3-a32b-d3e2b5658181

Registered design: 240 cells, 960 execution slots, and a maximum ceiling of 3,296,000 interactions. maximumExternalSpend is 0. The adapter remains diagnostic/unregistered, executionPermitted is false, resultsStatus is not_collected, and empiricalClaimPermitted is false.

This release records hypotheses, endpoints, scales, strata, estimands, missing-data rules, statistical rules, source lock, operation plan, expected manifest, results template, research decision ledger, data dictionary, and paper outline. It does not report an experiment, deployment, observed outcome, effectiveness claim, or comparative performance claim.