crewhaus v0.6.0

What's Changed

• docs: chvm is the recommended way to install crewhaus by @StudioMaxIO in #419
• chvm is at crewhaus/chvm, and covers Windows by @crewhaus in #420
• chvm installs in one command now by @crewhaus in #421
• fix(crew-orchestrator): forward per-role model_pool/model_tiers/model_fallbacks to the role loop by @crewhaus in #422
• feat(runtime-core): budget substrate — gate every model call, pool-aware degrade, toolsInPlay union, streaming model latency by @crewhaus in #423
• feat(trace): role/stage/profile attribution on model events, model_stage + model_directive kinds, durable routing event-log kinds by @crewhaus in #424
• feat(model-plan): new @crewhaus/model-plan package, strongest sentinel, requiresModelFeatures, adapter effectiveParams by @crewhaus in #425
• feat(eval-judge): judge and compaction spend on the run bus, budget.judge_share, run-spanning workflow meter by @crewhaus in #426
• feat(observability): OTel model attributes and metrics, escalation/judge/floor alerts and SLO, session pin/models fields, sessions tail and export by @crewhaus in #427
• feat(spec): models: profile registry, $profile references, hybrid pool blocks, escalate, judge panels, per-node/sub-agent routing by @crewhaus in #428
• feat(ir,compiler): IrModelProfile, $profile resolution at every slot, modelPlanIntegrity pass, capability/sunset validation by @crewhaus in #429
• feat(crew): per-candidate profile enrichment on roles and crew.routing.model for the llm router by @crewhaus in #431
• refactor(model-service): wireModels composition root wrapping today's emitter routing fragments byte-identically by @crewhaus in #430
• feat(tool-consult): Consult + Escalate tools, TrustOrigin consult, narrowRuleSet, interpreter parity by @crewhaus in #432
• feat(migration): MIGRATION_1_TO_2 via SpecEdits, irreversible guard, comment-preserving upgrade and migrate-all, --hoist-models, per-release upgrade notes, pricing-feed sunsets by @crewhaus in #433
• feat(runtime-core): per-candidate plan table — request params, subset tool advertisement, dispatch gate, per-candidate permissions/rate-limits/tool_config, volatile overlay by @crewhaus in #434
• feat(runtime-core): preRoute decision phase — directives with session pin, rules, capability eligibility, classifier hook, budget hints, pre-model mutate by @crewhaus in #435
• feat(model-service): guide, shadow, and committee side-call closures with nested single-turn isolation by @crewhaus in #437
• feat(sub-agents): per-sub-agent routing, projectParentHandle, Task profile allowlist, inherit_routing, child spend re-published by @crewhaus in #438
• feat(runtime-core): cascade and draft-verify — runOneTurn force, evaluation.on_fail escalate, clean-prompt snapshot, deferred quality fold by @crewhaus in #436
• feat(spec-patch): wildcard segment, model_pool structural rule, the 0.6.0 OPTIMIZABLE_PATHS delta with fixtures, Hangar spec-edit tiers, lint/doctor checks by @crewhaus in #439
• feat(model-router,routing-store): RouteSignals + hint input, policy union, scoped arms with backoff, per-candidate chains, v2 lines, priors, floor, route freeze by @crewhaus in #440
• feat(eval): routed evals with pinned seed and ephemeral scoreboard, served-model attribution, per-arm baselines, eval leaderboard by @crewhaus in #442
• feat(emitters): compiled bundles wire model-service runtime closures — Consult/Escalate, classifier, guide/shadow/committee — when the IR declares them by @crewhaus in #441
• feat(emitters): hybrid closure wiring on pipeline, research, batch and browser — closing the §11.3 shortfall by @crewhaus in #444
• feat(eval-grader): graders.yaml per_model, judge usage by (judge, arm), calibration by pair, judge calibrate --by-model by @crewhaus in #443
• feat(routing): reward.quality_source, route promote --gate, routing_promotion audit, feed-routing per stage by @crewhaus in #445
• feat(hangar): models tab and routes, per-role/profile cost aggregation, gutter kinds, advisor items and jobs, SECURITY_SURFACES by @crewhaus in #446
• feat(cli): models list|explain|audit|propose, route explain --json/status --by/propose, enumerateModelSlots, advise rules, routed from-advice, routing-aware canary, flywheel template, init --hybrid by @crewhaus in #447
• feat(eval-judge): wire the in-loop judge panel — judges/repeats/temperature/target at every judge site, and profile params on aux model slots by @crewhaus in #448
• fix(compiler): a profile's narrowing fields on a single-model slot are by design, not a pending row by @crewhaus in #449
• chore(release): crewhaus 0.6.0 by @crewhaus in #450
• docs(agents): AGENTS.md's routing pointer matched the shipped hybrid wiring by @crewhaus in #451

Full Changelog: v0.5.8...v0.6.0