Skip to content

enable-all-v8.md

Latest commit

 

History

History
133 lines (102 loc) · 4.13 KB

enable-all-v8.md

File metadata and controls

133 lines (102 loc) · 4.13 KB
Raw

Enable All v8 Features

This guide provides a single config profile that explicitly enables all major v8 feature families in openclaw-engram.

Apply under:

  • plugins.entries.openclaw-engram.enabled = true
  • plugins.entries.openclaw-engram.config = { ... }

Full v8 Config Profile

{
  "openaiApiKey": "${OPENAI_API_KEY}",
  "qmdEnabled": true,
  "qmdCollection": "openclaw-engram-hot-facts",
  "qmdColdTierEnabled": true,
  "qmdColdCollection": "openclaw-engram-cold",
  "conversationIndexEnabled": true,
  "conversationIndexBackend": "faiss",
  "conversationIndexFaissPythonBin": "python3",
  "conversationIndexFaissModelId": "text-embedding-3-small",
  "conversationIndexFaissIndexDir": "state/conversation-index/faiss",

  "recallPlannerEnabled": true,
  "memoryBoxesEnabled": true,
  "traceWeaverEnabled": true,
  "episodeNoteModeEnabled": true,
  "queryAwareIndexingEnabled": true,
  "multiGraphMemoryEnabled": true,
  "graphRecallEnabled": true,
  "graphAssistShadowEvalEnabled": true,
  "temporalMemoryTreeEnabled": true,

  "lifecyclePolicyEnabled": true,
  "lifecycleFilterStaleEnabled": true,
  "lifecycleMetricsEnabled": true,

  "procedural": {
    "enabled": true,
    "proceduralMiningCronAutoRegister": false
  },

  "proactiveExtractionEnabled": true,
  "contextCompressionActionsEnabled": true,
  "compressionGuidelineLearningEnabled": true,
  "compressionGuidelineSemanticRefinementEnabled": true,

  "identityEnabled": true,
  "identityContinuityEnabled": true,
  "continuityAuditEnabled": true,
  "continuityIncidentLoggingEnabled": true,

  "routingRulesEnabled": true,
  "sessionObserverEnabled": true,

  "sharedContextEnabled": true,
  "sharedCrossSignalSemanticEnabled": true,
  "compoundingEnabled": true,
  "compoundingInjectEnabled": true,
  "compoundingSemanticEnabled": true,
  "compoundingWeeklyCronEnabled": true,

  "qmdTierMigrationEnabled": true,
  "qmdTierAutoBackfillEnabled": true,

  "behaviorLoopAutoTuneEnabled": true,

  "debug": true
}

Safety Notes

  • Keep secrets in environment variables (${OPENAI_API_KEY}), not hardcoded keys.
  • If you run many features at once, expect higher extraction/consolidation activity.
  • debug: true is recommended while validating; disable later for quieter logs.
  • If you use conversationIndexBackend: "faiss", install scripts/faiss_requirements.txt first and optionally set ENGRAM_FAISS_ENABLE_ST=1 for sentence-transformers embeddings.
  • If you prefer QMD for transcript recall, swap the FAISS fields for conversationIndexBackend: "qmd" plus conversationIndexQmdCollection.
  • procedural.enabled turns on issue #519 procedural memory (writes under procedures/, recall injection, mining). The sample block above enables it; remove the procedural object or set "enabled": false if you want this profile without procedural behavior. See procedural-memory.md.

Required Restart

After config changes:

launchctl kickstart -k gui/$(id -u)/ai.openclaw.gateway

Verification Checklist

Run all commands:

openclaw engram compat --strict
openclaw engram stats
openclaw engram conversation-index-health
openclaw engram conversation-index-inspect
openclaw engram graph-health
openclaw engram tier-status
openclaw engram policy-status

Expected:

  • compat --strict: exits 0
  • stats: QMD: available
  • conversation-index-health: status: "ok" when backend is qmd
  • conversation-index-inspect: returns backend metadata and artifact state without mutating the index
  • graph-health: JSON report without runtime command failure
  • tier-status: returns migration telemetry JSON
  • policy-status: returns runtime policy snapshot JSON

Search Backend (v9.0)

The config above uses QMD (default). To use an alternative backend, add:

{
  "searchBackend": "orama"   // or "lancedb", "meilisearch", "remote", "noop"
}

See Search Backends for full options.

Related Docs