@joshuaswarren/openclaw-engram 9.0.17 → 9.0.19

package/README.md

@@ -30,7 +30,7 @@ AI agents forget everything between conversations. Engram fixes that.
 - **Local-first** — All memory data stays on your filesystem as plain markdown files. No cloud dependency, no vendor lock-in, fully portable.
 - **Pluggable search** — Choose from six search backends: QMD (hybrid BM25+vector+reranking), LanceDB, Meilisearch, Orama, remote HTTP, or bring your own.
 - **Memory OS features** — Graph recall, temporal memory tree, lifecycle policy, compounding, shared context, memory boxes, and identity continuity can be enabled progressively as your install grows.
-- **Benchmark-first roadmap** — Engram now has an evaluation harness with live shadow recall recording, so memory improvements can be measured on real agent trajectories instead of subjective recall demos.
+- **Benchmark-first roadmap** — Engram now has an evaluation harness with live shadow recall recording and a CI benchmark delta gate, so memory improvements can be measured and regression-checked instead of argued from anecdotes.
 - **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.
 
 ## Quick Start
@@ -139,7 +139,7 @@ Engram's capabilities are organized into feature families that you can enable pr
 | **Compounding** | Weekly synthesis that surfaces patterns and recurring mistakes |
 | **Hot/Cold Tiering** | Automatic migration of aging memories to cold storage |
 | **Behavior Loop Tuning** | Runtime self-tuning of extraction and recall parameters |
-| **Evaluation Harness** | Tracks benchmark packs, run summaries, and live shadow recall records so future PRs can be gated on memory quality instead of anecdotes |
+| **Evaluation Harness** | Tracks benchmark packs, run summaries, live shadow recall records, and CI delta comparisons so future PRs can be gated on memory quality instead of anecdotes |
 
 Start with defaults, then enable features as needed. See [Enable All Features](docs/enable-all-v8.md) for a full-feature config profile.
 
@@ -152,6 +152,8 @@ openclaw engram compat --strict              # Compatibility check
 openclaw engram benchmark-status             # Benchmark/eval harness packs, runs, shadow recalls, latest summaries
 openclaw engram benchmark-validate <path>    # Validate a benchmark manifest or pack directory
 openclaw engram benchmark-import <path>      # Import a validated benchmark pack into the eval store
+openclaw engram benchmark-ci-gate            # Compare base vs candidate eval stores and fail on regressions
+openclaw engram objective-state-status       # Objective-state snapshot counts and latest stored snapshot
 openclaw engram conversation-index-health    # Conversation index status
 openclaw engram graph-health                 # Entity graph status
 openclaw engram tier-status                  # Hot/cold tier metrics
@@ -174,6 +176,9 @@ Key settings:
 | `evalHarnessEnabled` | `false` | Enable the evaluation harness for benchmark packs, run summaries, and shadow recall bookkeeping |
 | `evalShadowModeEnabled` | `false` | Record live recall decisions to the eval store without changing injected output |
 | `evalStoreDir` | `{memoryDir}/state/evals` | Root directory for benchmark packs, run summaries, and shadow recall records |
+| `objectiveStateMemoryEnabled` | `false` | Enable the objective-state memory foundation for normalized world/tool state snapshots |
+| `objectiveStateSnapshotWritesEnabled` | `false` | Permit objective-state snapshot writers to persist typed state records |
+| `objectiveStateStoreDir` | `{memoryDir}/state/objective-state` | Root directory for objective-state snapshots |
 
 Full reference: [Config Reference](docs/config-reference.md)
 
@@ -183,7 +188,7 @@ Full reference: [Config Reference](docs/config-reference.md)
 - [Search Backends](docs/search-backends.md) — Choosing and configuring search engines
 - [Writing a Search Backend](docs/writing-a-search-backend.md) — Build your own adapter
 - [Config Reference](docs/config-reference.md) — Every setting with defaults
-- [Evaluation Harness](docs/evaluation-harness.md) — Benchmark pack and run-summary format
+- [Evaluation Harness](docs/evaluation-harness.md) — Benchmark pack, shadow recall, and CI delta gate format
 - [Architecture Overview](docs/architecture/overview.md) — System design and storage layout
 - [Retrieval Pipeline](docs/architecture/retrieval-pipeline.md) — How recall works
 - [Memory Lifecycle](docs/architecture/memory-lifecycle.md) — Write, consolidation, expiry