Files
Assistant/docs/PROJECT_STATUS.md

9.9 KiB
Raw Permalink Blame History

Assistant project status

Gen2 M6 checkpoint — 2026-08-30

  • GEN2-028 = COMPLETED_WITH_EVIDENCE: V1 compatibility inventory is recorded in docs/evidence/gen2-028-v1-compatibility-inventory-2026-08-30.json. Current production callers remain for PersonalMemory fallback, ActiveProjectFocus, truth_bundles, ordinary ENV reads, SkillProfile/FastPath; no path was proven SAFE_TO_RETIRE.
  • GEN2-029 = COMPLETED_WITH_SAFE_NOOP: no V1 runtime path or historical evidence was deleted or disabled. The decision and rollback point are recorded in docs/evidence/gen2-029-v1-retirement-decision-2026-08-30.json; FULL_RETIREMENT was not applicable because no irreversible retirement was performed.
  • GEN2-030_STATUS = ACCEPTANCE_IN_PROGRESS: this status is a documentation checkpoint, not a final Gen2 PASS. The final gate must confirm current production and repository evidence; the M6 retirement gap remains that active V1 compatibility callers have not yet been proven unused.
  • UI_BROWSER_ACCEPTANCE = DEFERRED_BY_OWNER; it is intentionally excluded from core/M6 acceptance and is neither counted as PASS nor as a blocker.
  • PROVIDER_GAPS = CPU_FAST_UNAVAILABLE; GPU_UNAVAILABLE; Local MEDIUM remains the primary route and Lifeboat remains fallback per existing M5 evidence.
  • GEN2_M6_FINAL_STATUS = NOT_FINAL_PASS_RETIREMENT_GAP_REMAINS; FULL_FINAL technical checks passed, but no final M6 baseline tag is created because active V1 compatibility callers were not proven safe to retire.
  • GEN2_M6_FINAL_EVIDENCE = docs/evidence/gen2-030-final-acceptance-2026-08-30.json.
  • GEN2_FINAL_STATUS = GEN2_CORE_ACCEPTED / V1_COMPATIBILITY_RETAINED / UI_DEFERRED.
  • GEN2-COMPAT-OBS-001 = INSTRUMENTATION_COMMITTED; PRODUCTION_OBSERVATION_BLOCKED_BY_MISSING_CAPABILITY; evidence: docs/evidence/gen2-compat-obs-001-blocked-2026-08-30.json. No V1 path is SAFE_TO_RETIRE; no retirement task was created.
  • `NEXT_EVOLUTION_QUEUE = instrument bounded compatibility usage, observe, then retire only individually proven paths under a future FULL_RETIREMENT gate; no Gen3 or new feature work is started.

Current audit delta — 2026-08-24

  • CURRENT_MODEL_SCOPE = LOCAL_FIRST_V1; one locally available Qwen3-8B or Gemma-4 conversational model is sufficient for the first stage, with the local Lifeboat as recovery fallback. External models remain supported as optional capability, but their absence must not block work: reflexes, skills and local executors provide degraded continuation. External/GPU qualification and promotion are intentionally deferred by DEV-000270.

  • TRUTH_FIRST_DEFAULT = LIVE; an unverified factual question is no longer passed to the advisory model as an answer. The response is explicitly marked verification_required; internal canonical status paths remain available. Semantic interpretation may select the provider-neutral WEB_RESEARCH capability without entity-specific keyword lists. Evidence: docs/evidence/truth-first-question-gate-2026-08-24-v1.json.

  • LAST_COMPLETED = DEV-000290; the governed self-development path, provider-neutral reachability rating, bounded authenticated operations smoke, truth-first research routing, relevance filtering and authenticated desktop/mobile chat acceptance are live. Deep mutation, delegation and long-running acceptance remain deferred.

  • CURRENT_PRODUCTION_STATUS = LIVE_HEALTHY; public /healthz is healthy and current runtime project labels match merged 8c6f23dc491150a747bcc46c26ff04d217d37eba. Evidence: docs/evidence/audit-completion-matrix-2026-08-24-v17.json.

  • FAST_PROVIDER_STATUS = LIVE_SMOKE_SUPPORTED: local CPU gateway authentication is configured through a protected file reference; four bounded local-only advisory cases completed through cognitive-worker. Evidence: docs/evidence/wp-601-fast-gateway-live-smoke-2026-08-22-v1.json.

  • MEDIUM_PROVIDER_STATUS = LIVE_BOUNDED_SMOKE_SUPPORTED_UNPROMOTED; four bounded local-provider cases passed with template thinking disabled. This is availability/response-shape evidence only; lifecycle qualification, activation and promotion remain unclaimed. Evidence: docs/evidence/medium-live-bounded-smoke-2026-08-23-v1.json.

  • HEAVY_PROVIDER_STATUS = LIVE_BOUNDED_INVOCATION_SUPPORTED_UNPROMOTED; the exact Domovoy adapter payload produced one bounded Qwen completion in 0.583s. Quality is not assessed, lifecycle promotion is not performed, and GPU availability is not treated as intelligence or routing priority. Evidence: docs/evidence/gpu-discovery-2026-08-23-v1.json, docs/evidence/gpu-invocation-qualification-2026-08-23-v1.json.

  • EXTERNAL_MODEL_STATUS = DEFERRED_BY_SCOPE; previous bounded GPU timeout evidence remains historical diagnostic evidence, but no external model is part of the current operating route and no further external probes are planned in this phase.

  • AGENT_EXECUTION_STATUS = LIVE_TYPED_GENERIC_LOCAL_BRIDGE; production worker and bridge are ready for task-evaluation and generic cognitive work. Generic output is recorded as an unverified Work Graph candidate and still requires independent synthesis before graph success. Evidence: docs/evidence/audit-completion-matrix-2026-08-24-v12.json.

  • ADDITIONAL_EVALUATION_STATUS = LIVE_TYPED_HANDOFF; Capability-owned unresolved evaluation decisions can now be handed to Work through a typed production endpoint. Work preserves parent evidence lineage and schedules the bounded additional profile through the existing durable graph/dispatch path; model/provider selection remains outside Work. Evidence: docs/evidence/additional-evaluation-runtime-live-2026-08-23-v1.json and DEV-000261.

  • FLEET_EXECUTION_EVIDENCE = docs/evidence/audit-completion-matrix-2026-08-24-v13.json; current live readiness has MEDIUM and LIFEBOAT ready, FAST and HEAVY unavailable by timeout, MEDIUM selected and LIFEBOAT fallback.

  • TRUTH_RESEARCH_STATUS = LIVE_CANARY_PERSISTED; the web research canary preserved an OBSERVED/COMPLETE/NO-CONFLICT source chain while leaving the claim UNVERIFIED and belief PROVISIONAL. Evidence: docs/evidence/truth-runtime-canary-2026-08-23-v1.json.

  • TRUTH_FIRST_CHAT_STATUS = LIVE_SOURCE_ROUTING_WITH_RELEVANCE_GATE; factual questions require strict semantic routing to optional WEB_RESEARCH; irrelevant search excerpts are rejected and the authenticated weather smoke returned an explicit no-confirmation response rather than invented data. Evidence: docs/evidence/audit-completion-matrix-2026-08-24-v15.json.

  • LEARNING_REFLECTION_STATUS = LIVE_RECONCILED; reflection worker returned no pending candidate and the durable state contains 31 RESOLVED candidates with no OPEN backlog. Evidence: docs/evidence/reflection-runtime-reconciliation-2026-08-23-v1.json.

  • BACKUP_RESTORE_STATUS = VERIFIED; latest checksum and isolated restore evidence is retained in docs/evidence/wp-002-postgres-restore-2026-08-22-v1.json.

  • GOVERNED_SELF_DEVELOPMENT_STATUS = DURABLE_GATE_VERIFIER_AND_TRANSITIONS_LIVE; Work-owned revisioned gate storage, independent local verification, and authenticated optimistic-concurrency transitions for canary/promotion/rollback are live. Source mutation and external bridge execution remain unclaimed. Evidence: docs/evidence/governed-development-gate-persistence-2026-08-23-v1.json and DEV-000254 CI/live-smoke records.

  • UI_AUTHENTICATED_ACCEPTANCE = AUTHENTICATED_QUICK_SMOKE_PASS; browser-driven smoke completed in the active production session: chat response received, admin login accepted, diagnostics 7/7 passed, operations fleet 4/4 READY, and action history loaded. Deep mutation, delegation-specific and long-running acceptance remain intentionally deferred.

  • CURRENT_UI_SESSION_STATUS = AUTHENTICATED_DESKTOP_MOBILE_PASS; the current signed-in session showed active state and no horizontal overflow at 390×844 after the chat layout/header fixes. Deep mutation/delegation/long-running acceptance remains deferred.

  • AGENT_EXECUTION_BRIDGE_STATUS = LIVE_LOCAL_TYPED_BRIDGE; the independent local bridge reports healthz=200 with task-evaluation and generic-cognitive-work; agent-worker reports mode=execution, executor_configured=true, ready=true. Evidence: docs/evidence/audit-completion-matrix-2026-08-24-v13.json.

  • FLEET_READINESS_LIVE = MEDIUM_SELECTED_BY_AVAILABILITY; the latest live readiness endpoint observed FAST unavailable, MEDIUM available, HEAVY unavailable by timeout and LIFEBOAT available; it selected MEDIUM while the configured preference remained FAST. This is routing evidence only, not a quality or promotion claim. Evidence: docs/evidence/audit-completion-matrix-2026-08-24-v13.json.

  • UI_AUTHENTICATED_ACCEPTANCE = DEEP_DESKTOP_PASS_WITH_MOBILE_GAP; deep signed-in browser evidence covers chat reply, diagnostics 7/7 after refresh, fleet/agent/action-history refresh, 3/4 fleet readiness with FAST selected, truthful local executor state and learning/reflection refresh. Static responsive regression guard passed in CI; mobile viewport acceptance is not claimed. Evidence: docs/evidence/deep-authenticated-acceptance-2026-08-23-v1.json.

Current governance checkpoint

  • COMPLETION_MATRIX = docs/evidence/audit-completion-matrix-2026-08-24-v17.json; generic candidate envelope verification, independent Work Graph synthesis, research relevance gating and authenticated desktop/mobile truth-first chat acceptance are live; external qualification and deep mutation/delegation/long-running acceptance remain explicit.

  • ACTIVE_TASK = dev/tasks/DEV-000291-mobile-acceptance-evidence.json revision 1; this evidence reconciliation does not change runtime behavior or model routing.

  • LINEAGE_DECISION = docs/GOVERNANCE_LINEAGE_DECISION_2026-08-22.md; historical packets remain provenance and are not current authority.

  • REGISTRY_STATUS = YAML and JSON semantic mirrors agree

  • PRODUCTION_DOMAIN = docker.cin.su; shared docker-test is not used for this project.

  • NEXT_WORK = obtain a viewport-capable browser for real mobile authenticated visual acceptance; add external providers later under an explicit task. Deep UI/load tests remain deferred until explicitly requested.