@joshuaswarren/openclaw-engram 9.0.47 → 9.0.49

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -47,6 +47,7 @@ AI agents forget everything between conversations. Engram fixes that.
47
47
  - **Artifact recovery recall** — Engram can now, when both `creationMemoryEnabled` and `workProductRecallEnabled` are enabled, surface prompt-relevant work-product ledger entries back into recall as a dedicated `Work Products` section and inspect reuse candidates with `openclaw engram work-product-recall-search`.
48
48
  - **Commitment lifecycle foundation** — Engram can now, when `creationMemoryEnabled`, `commitmentLedgerEnabled`, and `commitmentLifecycleEnabled` are enabled, transition existing commitments to `fulfilled` / `cancelled` / `expired`, inspect overdue and stale obligations in `openclaw engram commitment-status`, and run deterministic lifecycle cleanup with `openclaw engram commitment-lifecycle-run`.
49
49
  - **Resume-bundle builder** — Engram can now, when `creationMemoryEnabled` and `resumeBundlesEnabled` are enabled, persist typed crash-recovery resume bundles, inspect them with `openclaw engram resume-bundle-status`, write explicit handoff shells through `openclaw engram resume-bundle-record`, and build bounded resume bundles from transcript recovery, recent objective-state snapshots, work products, and open commitments through `openclaw engram resume-bundle-build`.
50
+ - **Utility-learning offline learner** — Engram can now, when `memoryUtilityLearningEnabled` is enabled, persist typed downstream utility events for promotion and ranking decisions, inspect the raw event ledger with `openclaw engram utility-status`, record deterministic benchmark/operator telemetry through `openclaw engram utility-record`, and learn bounded offline promotion/ranking weights through `openclaw engram utility-learn` and `openclaw engram utility-learning-status` before runtime weighting lands.
50
51
  - **Zero-config start** — Install, add an API key, restart. Engram works out of the box with sensible defaults and progressively unlocks advanced features as you enable them.
51
52
 
52
53
  ## Quick Start
@@ -183,6 +184,10 @@ openclaw engram commitment-lifecycle-run # Expire overdue commitments and clean
183
184
  openclaw engram work-product-status # Work-product ledger counts and latest recorded output
184
185
  openclaw engram work-product-record # Record a typed work-product ledger entry
185
186
  openclaw engram work-product-recall-search <query> # Preview reusable work products from the creation-memory ledger
187
+ openclaw engram utility-status # Utility-learning telemetry counts and latest observed outcome event
188
+ openclaw engram utility-record # Record a typed utility-learning telemetry event
189
+ openclaw engram utility-learning-status # Latest offline utility-learning snapshot and learned weight counts
190
+ openclaw engram utility-learn # Learn bounded offline promotion/ranking weights from recorded utility events
186
191
  openclaw engram conversation-index-health # Conversation index status
187
192
  openclaw engram graph-health # Entity graph status
188
193
  openclaw engram tier-status # Hot/cold tier metrics
@@ -225,6 +230,8 @@ Key settings:
225
230
  | `semanticRulePromotionEnabled` | `false` | Enable deterministic promotion of explicit `IF ... THEN ...` rules from verified episodic memories via `openclaw engram semantic-rule-promote` |
226
231
  | `semanticRuleVerificationEnabled` | `false` | Verify promoted semantic rules against their cited source episodes at recall time and inject a dedicated `Verified Rules` section via `openclaw engram semantic-rule-verify` |
227
232
  | `creationMemoryEnabled` | `false` | Enable the creation-memory foundation, including the work-product ledger and operator-facing write/status commands |
233
+ | `memoryUtilityLearningEnabled` | `false` | Enable typed utility-learning telemetry storage for future promotion/ranking learners |
234
+ | `promotionByOutcomeEnabled` | `false` | Reserve outcome-aware promotion controls that later slices will drive from learned utility signals |
228
235
  | `commitmentLedgerEnabled` | `false` | Enable the explicit commitment ledger for promises, follow-ups, deadlines, and unfinished obligations |
229
236
  | `commitmentLifecycleEnabled` | `false` | Enable commitment lifecycle transitions, stale tracking, and resolved-entry cleanup for the commitment ledger |
230
237
  | `commitmentStaleDays` | `14` | Days before an open commitment without a due date is considered stale in lifecycle status |