Skip to content
← Back to changelog index

17.2.5

Oct 27, 2026 · 14 packages · 16 unique changes · 8 release entries

appsbundlesintegrationslibsmodulesBreaking changes

This release affects the contracts, integrations, sharedLibs, solutions families.

Run contractspec connect adoption resolve --family contracts to see how it impacts your project.

Release summaries

  • agent-host-extraction-lane-x8

    Agent Host protocol, automation, model-route, lifecycle and Runtime Node V3 execution behavior moves to its platform owners

    integrator

    New explicit subpaths expose the governed automation run lifecycle, model qualification/canary head transitions, model route V2 admission, the application qualification port and agent export artifacts (ai-agent); the application installation lifecycle, application qualification contracts/validation and the markdown migration material schema (agent-protocol-spec); the PostgreSQL application lifecycle adapter and migration (lssm-os-state-postgres); and the Runtime Node V3 execution service (lssm-os-runtime).

  • agent-operations-harness-ownership

    Give agent operations and Harness contracts their canonical runtime owners

    integrator

    Legacy imports preserve operation schemas and the same scenario and suite registry classes through canonical owners.

  • ai-canonical-runtime-owners

    Preserve complete AI runtime bodies through canonical ownership imports.

    maintainer

    Preserve provider factory signatures, policy gates, nominal registry identity and complete runtime defaults.

  • ai-improvement-v2-human-promotion

    Governed V2 AI improvement work orders reject failed evaluations and promote to production only with an explicit human approval; policy approvals and auto-apply policies stop after canary.

    integrator

    Breaking safety correction for V2 only. Previously a failed evaluation waited for approval and any approval (including approverType policy) or a qualified auto-apply policy could promote it. Now a failed evaluation ends as rejected with failure code evaluation_failed, a policy approval covers canary only, and a healthy canary without a human approval waits at awaiting_approval. Approving it resumes observation and then promotes. The V1 runtime and the receipt binding digest are unchanged; the additive AiImprovementApprovalStageV2 type and optional receipt stage field are new. Major bump rationale (semver for a breaking change, compatibility-versioning rule): the removed behaviour violated the documented no-autonomous-promotion policy and has no in-repository production consumer.

  • ai-protocol-v2-clock-boundary

    Preserve historical optional-clock model validation at its runtime owner.

    maintainer

    Preserve guarded lazy clock reads, field-read order, issue order and fallback short-circuiting through complete structural phases.

  • cmp-agent-epochs-host-composition

    Agent Host composes the bounded-epoch runner for ContractSpecAgent runs, and an open epoch's token and tool-call spend survives a restart

    integrator

    runAgentBoundedEpochsV1 stores the open epoch's token and tool-call spend, with the number of steps it counts and the number of steps that had reached a journaled effect (openEpochUsage), with its checkpoint. A restarted process continues those counters, so a crash-looping epoch exhausts its budget instead of getting fresh counters; it runs the admitted steps again without a budget check, so each replays its recorded effects or stops the run at its pending one, even after the wall-time budget ran out, and a journaled step is never sent again in a later epoch; and a restart past maxResumes closes the epoch with the persisted spend. An epoch that ends on its budget or a cancellation with a pending journal entry stops as effect_indeterminate instead of renewing, and the result's optional stopDetail names the effect. Steps keep reporting their spend as before: one whose effects all replay is counted once. Records stored before the change load unchanged and resume their open epoch from zero. Agent Host adds createAgentHostBoundedAgentRunnerV1, which bounds a ContractSpecAgent run by a host-owned epoch policy instead of maxSteps alone and journals every generate call, plus the optional boundedAgentRuns server binding and bounded_agent_runs readiness row.

    maintainer

    The durable record schema accepts an optional openEpochUsage (tokens, toolCalls, steps, admittedSteps) that must belong to the open epoch_started checkpoint. executeAgentEpochV1 takes the optional spent, onAdmit and onSpend(spent, durable) inputs: steps below spent.admittedSteps run without the budget check or the wall-time abort, onAdmit fires when a step first calls context.effect so its first journal write carries the admission, a step below spent.steps whose effects all replayed is not counted again, and durable is false when the step ran a journaled effect, so the next write carries its spend (the runner always stores it before the boundary decision). An execution that ends budget_exhausted or cancelled with a pending journal entry ends as effect_indeterminate, with a detail the runner returns as stopDetail. The Agent Host step runs each generate call through context.effect under the key agent-host:generate:<epoch>:<step>, reports the recorded spend of a replayed call, and turns any failure or abort of a call whose journal entry is written into effect_indeterminate, naming the abort reason or the error's class name. The bounded-epoch DocBlock and the architecture-invariants ledger row describe the persisted spend and the remaining gaps.

  • headless-chat-runtime

    Move canonical headless chat implementation to the lower AI runtime while preserving module compatibility.

    integrator

    Headless consumers may use the additive AI runtime chat entry; existing module imports keep their canonical identities.

  • refactor-move-dev-health-repairs

    Repair relocation regressions in locale imports, Core type imports, workspace ranges and repository metadata

    integrator

    Relocated packages resolve their locale helpers and Core types again and publish with accurate repository provenance metadata.

Migration guide

  • Optionally run ContractSpecAgent loops through the Agent Host bounded-epoch runner

    Additive. The bounded_agent_runs row stays disabled and nothing changes until an owner binds a policy.

    When: When an Agent Host installation runs ContractSpecAgent loops that can outlive one budget window or one process.

    1. Apply migratePostgresAgentDurableStore from @lssm-tech/integration.agent-durable-store/postgres/migrations to the runtime database (pass the same schema and table options as the binding's tables).
    2. Supply the boundedAgentRuns server binding with an agent id and an agentBoundedEpochPolicyV1Schema policy sized for the task; an approval_required policy also needs approveRenewal.
    3. Start runs with duties.boundedAgentRuns.run and a runRef that is stable across restarts; reconcile any run that stops as effect_indeterminate before starting it again under a new runRef.
  • Keep reporting step spend when an effect replays

    No change is required. Steps report their spend as before; a step that only replays journaled effects may report their recorded spend, and the runner counts it once (it does not count again a step that the restored counters already hold). A restored step that also runs a new effect is counted in full. Report the real spend of work done outside context.effect, which a restart runs again.

    When: When a custom AgentEpochStepV1 wraps context.effect and runs under runAgentBoundedEpochsV1.

    1. Report the recorded spend of a replayed effect, or 0; do not add a replay-detection flag to the step.
  • Expect effect_indeterminate when an epoch ends with a pending effect

    Fail-closed change. An epoch whose step was aborted (wall time or cancellation) while one of its effects was still pending used to end as budget_exhausted or cancelled and could renew; it now stops as effect_indeterminate, because a renewed epoch would run that work again under new keys.

    When: When a caller of runAgentBoundedEpochsV1 branches on stopReason or retries runs that ended budget_exhausted or cancelled.

    1. Treat effect_indeterminate as a run to reconcile (read stopDetail for the effect key) before starting the work again under a new runRef.
  • Optionally import the moved behavior from its owner subpaths

    Additive only; the agent-host root and ./t1 keep every previous name as a named alias, and existing package exports are unchanged. lib.ai-agent gains a lib.agent-evals dependency that only the model-route subpaths load.

    1. Replace agent-host imports with the owner subpaths, for example createRuntimeNodeV3ExecutionServiceV1 from @lssm-tech/lib.lssm-os-runtime/runtime-node/v3-execution-service.
    2. Bind AgentAutomationAuthorityPortV1<TApplication> and AgentAutomationMaterializeInputV1<TApplication> to the host's materialized application type.
  • Supply a human approval before V2 production promotion

    Required

    Auto-apply and policy approvals no longer promote V2 work orders to production.

    When: When a V2 work order reaches awaiting_approval after a healthy canary observation.

    1. Route the awaiting work order to a human reviewer.
    2. Submit an approval with approverType human bound to the current work order digest.
    3. Run the work item again; it re-observes the canary and then promotes.
  • Treat failed evaluations as terminal

    Required

    A V2 evaluation with testsPassed false now ends as rejected (evaluation_failed).

    When: When consumers previously approved a failed evaluation.

    1. Enqueue a new proposal instead of approving the failed work order.

Upgrade steps

  • Verify persisted in-epoch usage and the host runner

    auto

    Run the epoch runtime tests and the Agent Host bounded agent-run tests.

    Packages: @lssm-tech/lib.ai-agent, @lssm-tech/lib.agent-protocol-spec, @lssm-tech/app.agent-host

    1. Run bun test src/epochs src/agent/bounded-epoch-step.test.ts in ai-agent (usage.test.ts covers restarts, W5 records, refused usage and budget-checked unjournaled steps; replay-region.test.ts covers crashes before epoch_ended, counting a replay once and restarts after the wall-time budget; run.test.ts covers the pending-effect stop).
    2. Run bun test ./src/bounded-agent-run.test.ts ./src/bounded-agent-run-recovery.test.ts ./src/bounded-agent-run-duty.test.ts ./src/bounded-epoch-recovery.test.ts in agent-host.
  • Mark automated approvals as canary-only

    assisted

    Policy approvers may set stage canary; stage promotion is refused for non-human approvers.

    Packages: @lssm-tech/lib.ai-agent

    1. Keep approverType policy for automated canary gating.
    2. Never issue stage promotion from an automated approver.

Unique release changes

  • Repair relocation regressions in locale imports, Core type imports, workspace ranges and repository metadata

    6 packages · 6 occurrences

  • Additive only; the agent-host root and ./t1 keep every previous name as a named alias, and existing package exports are unchanged. lib.ai-agent gains a lib.agent-evals dependency that only the model-route subpaths load.

    4 packages · 4 occurrences

  • Agent Host protocol, automation, model-route, lifecycle and Runtime Node V3 execution behavior moves to its platform owners

    4 packages · 4 occurrences

  • Additive. The bounded_agent_runs row stays disabled and nothing changes until an owner binds a policy.

    3 packages · 3 occurrences

  • Agent Host composes the bounded-epoch runner for ContractSpecAgent runs, and an open epoch's token and tool-call spend survives a restart

    3 packages · 3 occurrences

  • Fail-closed change. An epoch whose step was aborted (wall time or cancellation) while one of its effects was still pending used to end as budget_exhausted or cancelled and could renew; it now stops as effect_indeterminate, because a renewed epoch would run that work again under new keys.

    3 packages · 3 occurrences

  • Give agent operations and Harness contracts their canonical runtime owners

    3 packages · 3 occurrences

  • Move canonical headless chat implementation to the lower AI runtime while preserving module compatibility.

    3 packages · 3 occurrences

  • No change is required. Steps report their spend as before; a step that only replays journaled effects may report their recorded spend, and the runner counts it once (it does not count again a step that the restored counters already hold). A restored step that also runs a new effect is counted in full. Report the real spend of work done outside context.effect, which a restart runs again.

    3 packages · 3 occurrences

  • Run the epoch runtime tests and the Agent Host bounded agent-run tests.

    3 packages · 3 occurrences

  • Preserve complete AI runtime bodies through canonical ownership imports.

    2 packages · 2 occurrences

  • A V2 evaluation with testsPassed false now ends as rejected (evaluation_failed).

    1 package · 1 occurrence

  • Auto-apply and policy approvals no longer promote V2 work orders to production.

    1 package · 1 occurrence

  • Governed V2 AI improvement work orders reject failed evaluations and promote to production only with an explicit human approval; policy approvals and auto-apply policies stop after canary.

    1 package · 1 occurrence

  • Policy approvers may set stage canary; stage promotion is refused for non-human approvers.

    1 package · 1 occurrence

  • Preserve historical optional-clock model validation at its runtime owner.

    1 package · 1 occurrence

Impacted packages

  • @lssm-tech/app.agent-host

    Layer: apps · 5 changes

  • @lssm-tech/bundle.workspace

    Layer: bundles · 1 change

  • @lssm-tech/integration.cli-contractspec-ai

    Layer: integrations · 1 change

  • @lssm-tech/integration.cli-contractspec-builder

    Layer: integrations · 1 change

  • @lssm-tech/integration.lssm-os-state-postgres

    Layer: integrations · 2 changes

  • @lssm-tech/lib.agent-protocol-spec

    Layer: libs · 2 changes

  • @lssm-tech/lib.ai-agent

    Layer: libs · 1 change

  • @lssm-tech/lib.contracts-spec

    Layer: libs · 1 change

  • @lssm-tech/lib.contracts-spec-core

    Layer: libs · 1 change

  • @lssm-tech/lib.forms-runtime

    Layer: libs · 1 change

  • @lssm-tech/lib.harness

    Layer: libs · 1 change

  • @lssm-tech/lib.lssm-os-runtime

    Layer: libs · 2 changes

  • @lssm-tech/lib.workflow-composer

    Layer: libs · 1 change

  • @lssm-tech/module.ai-chat

    Layer: modules · 1 change