@xdarkicex/openclaw-memory-libravdb 1.4.30 → 1.4.32

package/README.md

@@ -12,22 +12,44 @@
 </div>
 
 `@xdarkicex/openclaw-memory-libravdb` is a local-first OpenClaw memory plugin
-backed by the `libravdbd` daemon. It replaces the lightweight default memory
+backed by the `libravdbd` vector service. It replaces the lightweight default memory
 path with scoped session, user, and global memory; continuity-aware prompt
 assembly; durable recall; and sidecar-owned compaction.
 
 [Install](./docs/install.md) · [Full installation reference](./docs/installation.md) · [Architecture](./docs/architecture.md) · [Security](./docs/security.md) · [Performance and tuning](./docs/performance-and-tuning.md) · [Contributing](./docs/contributing.md)
 
-New install? Start here: [Install guide](./docs/install.md). Preferred setup on
-macOS: install `libravdbd` with Homebrew, install the OpenClaw plugin, then
-assign the plugin to the OpenClaw memory slot.
+New install? Start here: [Install guide](./docs/install.md).
 
 ## Install
 
+Install `libravdbd` with your system package manager, then install
+the OpenClaw plugin.
+
+**macOS (Homebrew)**
+
 ```bash
 brew tap xDarkicex/homebrew-openclaw-libravdb-memory
 brew install libravdbd
 brew services start libravdbd
+```
+
+**Linux (APT)**
+
+```bash
+sudo apt install libravdbd
+systemctl --user enable --now libravdbd
+```
+
+**Linux (AUR)**
+
+```bash
+yay -S libravdbd-bin
+systemctl --user enable --now libravdbd
+```
+
+**Plugin (all platforms)**
+
+```bash
 openclaw plugins install @xdarkicex/openclaw-memory-libravdb
 ```
 
@@ -51,8 +73,11 @@ Then activate the plugin in `~/.openclaw/openclaw.json`:
 }
 ```
 
-Verify the daemon and plugin:
+Verify the service and plugin:
 
+```bash
+openclaw memory status
+```
 
 Healthy output should show `Sidecar=running`, stored memory counts, the active
 gate threshold, and the loaded embedding profile.
@@ -63,7 +88,7 @@ Runtime requirements:
 
 - OpenClaw `>= 2026.3.22`
 - Node.js `>= 22`
-- a separately installed `libravdbd` daemon
+- a separately installed `libravdbd` service
 
 Compatibility note:
 
@@ -71,11 +96,11 @@ Compatibility note:
 
 Default endpoints:
 
-- macOS/Linux user-local daemon: `unix:$HOME/.clawdb/run/libravdb.sock`
-- Homebrew daemon on Apple Silicon: `unix:/opt/homebrew/var/clawdb/run/libravdb.sock`
-- Windows daemon: `tcp:127.0.0.1:37421`
+- macOS/Linux user-local service: `unix:$HOME/.libravdbd/run/libravdb.sock`
+- Homebrew service on Apple Silicon: `unix:/opt/homebrew/var/libravdbd/run/libravdb.sock`
+- Windows service: `tcp:127.0.0.1:37421`
 
-If your daemon runs elsewhere, set `sidecarPath`:
+If your service runs elsewhere, set `sidecarPath`:
 
 ```json
 {
@@ -108,14 +133,14 @@ If your daemon runs elsewhere, set `sidecarPath`:
   newest working context.
 - **Local-first inference** - uses local embedding and compaction paths by
   default, with optional external summarizer configuration.
-- **Explicit daemon lifecycle** - the npm/OpenClaw package stays connect-only;
+- **Explicit service lifecycle** - the npm/OpenClaw package stays connect-only;
   `libravdbd` is installed and supervised separately.
 
 ## Security Defaults
 
 Stored memory is treated as untrusted historical context. Retrieved memory is
 framed before it reaches the downstream model, memory collections are scoped by
-session/user/global namespace, and daemon installation is outside the npm plugin
+session/user/global namespace, and service installation is outside the npm plugin
 package.
 
 Before exposing OpenClaw over remote channels, read [Security](./docs/security.md).
@@ -123,7 +148,7 @@ Before exposing OpenClaw over remote channels, read [Security](./docs/security.m
 ## Operator Quick Refs
 
 ```bash
-openclaw memory status
+openclaw memory status [--deep] [--json]
 openclaw memory index --force
 openclaw memory search "prior context"
 openclaw memory export --user-id <userId>
@@ -132,17 +157,142 @@ openclaw memory journal --limit 50
 openclaw memory dream-promote --user-id <userId> --dream-file /path/to/DREAMS.md
 ```
 
-Use [Install](./docs/install.md) for daemon lifecycle commands and
+Use [Install](./docs/install.md) for service lifecycle commands and
 [Uninstall](./docs/uninstall.md) for safe shutdown and removal.
 
+## Configuration Reference
+
+All configuration keys are optional.
+
+If you do not have a GPU available, setting `onnxDevice` to `"cpu"` is
+recommended to avoid startup failures from missing GPU/NPU providers,
+but this is also optional — the service auto-detects and falls back to
+CPU when a provider is unavailable.
+
+### Connection
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `sidecarPath` | string | `auto` | `"auto"` probes standard socket paths; set `unix:/path` or `tcp:host:port` to override |
+| `grpcEndpoint` | string | — | Optional gRPC kernel endpoint for hosts using the gRPC kernel transport |
+| `rpcTimeoutMs` | number | `30000` | Per-call timeout for service RPC (ms) |
+| `dbPath` | string | auto-named | Explicit DB path; when set bypasses model-specific naming |
+
+### Embedding
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `embeddingProfile` | string | `nomic-embed-text-v1.5` | Primary embedding model |
+| `fallbackProfile` | string | `bge-small-en-v1.5` | Fallback when primary model fails dimension checks |
+| `embeddingBackend` | string | — | `bundled`, `onnx-local`, or `custom-local` |
+| `onnxDevice` | string | `auto` | ONNX execution provider: `auto`, `cpu`, `coreml` (macOS), `cuda` (Linux/Windows), `directml` (Windows), `openvino` (Linux) |
+| `embeddingRuntimePath` | string | — | Path to ONNX runtime library (maps to `LIBRAVDB_ONNX_RUNTIME`) |
+| `embeddingModelPath` | string | — | Path to custom embedding model `.onnx` file |
+| `embeddingTokenizerPath` | string | — | Path to custom tokenizer file |
+| `embeddingDimensions` | number | — | Embedding dimension override |
+| `embeddingNormalize` | boolean | — | Enable embedding normalization |
+
+### Retrieval
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `topK` | number | — | Max results per search |
+| `alpha` | number | — | Semantic similarity weight |
+| `beta` | number | — | Recency weight |
+| `gamma` | number | — | Summary quality weight |
+| `crossSessionRecall` | boolean | `true` | When `false`, only session-scoped memories are retrieved |
+| `useSessionRecallProjection` | boolean | — | Use `session_recall` collection instead of `session` |
+| `useSessionSummarySearchExperiment` | boolean | — | Use `session_summary` collection for search |
+
+### Ingestion gating
+
+Gating thresholds and scoring weights are owned by the vector service and configured via
+service environment variables. See the service documentation for tuning details.
+
+The plugin exposes `ingestionGateThreshold` for host-side gating decisions:
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `ingestionGateThreshold` | number | `0.35` | Minimum semantic relevance score used by the plugin host for ingestion gating |
+
+### Compaction
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `compactThreshold` | number | — | Absolute token threshold for forced compaction |
+| `compactionThresholdFraction` | number | `0.8` | Dynamic trigger as fraction of active token budget |
+| `compactSessionTokenBudget` | number | `2000` | Auto-compact when session exceeds this many tokens since last compaction; `0` disables |
+| `compactionQualityWeight` | number | `0.5` | How much summary confidence affects retrieval score (0 = ignore, 1 = full suppression) |
+
+### Summarizer
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `summarizerBackend` | string | — | `bundled`, `onnx-local`, `ollama-local`, or `custom-local` |
+| `summarizerProfile` | string | — | Summarizer model profile |
+| `summarizerModel` | string | — | Model name for summarization |
+| `summarizerModelPath` | string | — | Path to summarizer model file |
+| `summarizerTokenizerPath` | string | — | Path to summarizer tokenizer file |
+| `summarizerRuntimePath` | string | — | Path to summarizer ONNX runtime |
+| `summarizerEndpoint` | string | — | External summarizer endpoint URL |
+| `ollamaUrl` | string | — | Ollama base URL (populates `summarizerEndpoint` when unset) |
+| `compactModel` | string | — | Model to use for compaction summaries (populates `summarizerModel` when unset) |
+
+### Identity
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `userId` | string | auto-derived | Stable user identity for cross-session durable memory |
+| `identityPath` | string | `$OPENCLAW_STATE_DIR/libravdb-identity.json` | Custom path for the auto-derived identity file |
+
+### Markdown ingestion
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `markdownIngestionEnabled` | boolean | `false` | Watch markdown roots for changes |
+| `markdownIngestionRoots` | string[] | — | Directories to watch |
+| `markdownIngestionInclude` | string[] | — | Glob patterns to include |
+| `markdownIngestionExclude` | string[] | — | Glob patterns to exclude |
+| `markdownIngestionDebounceMs` | number | `150` | Debounce window for file change events |
+| `markdownIngestionObsidianEnabled` | boolean | `false` | Watch Obsidian vault roots |
+| `markdownIngestionObsidianRoots` | string[] | — | Obsidian vault directories |
+| `markdownIngestionObsidianInclude` | string[] | — | Obsidian glob include patterns |
+| `markdownIngestionObsidianExclude` | string[] | — | Obsidian glob exclude patterns |
+| `markdownIngestionObsidianDebounceMs` | number | `150` | Obsidian debounce window |
+
+### Dream promotion
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `dreamPromotionEnabled` | boolean | `false` | Enable dream diary promotion |
+| `dreamPromotionDiaryPath` | string | — | Path to dream diary markdown file |
+| `dreamPromotionUserId` | string | — | User ID for dream collection scoping |
+| `dreamPromotionDebounceMs` | number | `150` | Debounce window for dream diary changes |
+
+### Misc
+
+| Key | Type | Default | Notes |
+|---|---|---|---|
+| `sessionTTL` | number | — | Session TTL in seconds |
+| `recencyLambdaSession` | number | — | Session recency decay factor |
+| `recencyLambdaUser` | number | — | User recency decay factor |
+| `recencyLambdaGlobal` | number | — | Global recency decay factor |
+| `tokenBudgetFraction` | number | — | Fraction of host token budget to use for memory context |
+| `maxRetries` | number | — | Max RPC retries |
+| `logLevel` | string | — | Log level override |
+| `lifecycleJournalMaxEntries` | number | `500` | Max lifecycle journal entries |
+| `authoredHardBudgetFraction` | number | — | Token budget fraction for hard-authored recall (0–1) |
+| `authoredSoftBudgetFraction` | number | — | Token budget fraction for soft-authored recall (0–1) |
+| `elevatedGuidanceBudgetFraction` | number | — | Token budget fraction for elevated guidance recall (0–1) |
+
 ## Optional Features
 
 - **Markdown ingestion** watches OpenClaw-owned markdown roots or Obsidian vaults
   and syncs eligible notes into memory. See [Features](./docs/features.md).
 - **Dream promotion** promotes vetted dream diary bullets into an isolated
   `dream:{userId}` collection. See [Features](./docs/features.md).
-- **Embedding profiles** expose local model metadata defaults for MiniLM and
-  Nomic. See [Embedding profiles](./docs/embedding-profiles.md).
+- **Embedding profiles** default to `nomic-embed-text-v1.5` with `bge-small-en-v1.5`
+  fallback. See [Embedding profiles](./docs/embedding-profiles.md).
 
 ## Docs By Goal
 
@@ -162,8 +312,8 @@ bash scripts/build-daemon.sh
 ```
 
 `scripts/build-daemon.sh` prepares `.daemon-bin/libravdbd` for local plugin
-testing when you have a published daemon binary, a Homebrew daemon, or a local
-daemon checkout. For the full source workflow, read [Development](./docs/development.md).
+testing when you have a published service binary, a Homebrew service, or a local
+service checkout. For the full source workflow, read [Development](./docs/development.md).
 
 ## Runtime Facts
 
@@ -171,6 +321,6 @@ daemon checkout. For the full source workflow, read [Development](./docs/develop
 - OpenClaw plugin id: `libravdb-memory`
 - plugin kind: `memory`, `context-engine`
 - minimum OpenClaw host version: `>= 2026.3.22`
-- default data path: `$HOME/.clawdb/data.libravdb`
-- default macOS/Linux endpoint: `unix:$HOME/.clawdb/run/libravdb.sock`
+- default data path: `$HOME/.libravdbd/data_nomic-embed-text-v1_5.libravdb`
+- default macOS/Linux endpoint: `unix:$HOME/.libravdbd/run/libravdb.sock`
 - default Windows endpoint: `tcp:127.0.0.1:37421`

package/dist/sidecar.js

@@ -342,11 +342,11 @@ export function defaultEndpoint(platform = process.platform, homeDir = os.homedi
     const sockName = "libravdb.sock";
     const candidateDirs = [
         // User-local (npm plugin convention)
-        homeDir?.trim() ? path.join(homeDir, ".clawdb", "run") : null,
+        homeDir?.trim() ? path.join(homeDir, ".libravdbd", "run") : null,
         // Homebrew (Apple Silicon) — matches the Homebrew formula LaunchAgent
-        "/opt/homebrew/var/clawdb/run",
+        "/opt/homebrew/var/libravdbd/run",
         // Homebrew (Intel Mac) / manual Linux installs
-        "/usr/local/var/clawdb/run",
+        "/usr/local/var/libravdbd/run",
     ].filter((d) => d !== null);
     for (const dir of candidateDirs) {
         const sockPath = path.join(dir, sockName);
@@ -361,8 +361,8 @@ export function defaultEndpoint(platform = process.platform, homeDir = os.homedi
     }
     // Fallback to the original user-local path so error messages stay familiar.
     const baseDir = homeDir?.trim()
-        ? path.join(homeDir, ".clawdb", "run")
-        : path.join(".", ".clawdb", "run");
+        ? path.join(homeDir, ".libravdbd", "run")
+        : path.join(".", ".libravdbd", "run");
     return `unix:${path.join(baseDir, sockName)}`;
 }
 export function buildSidecarEnv(cfg) {
@@ -424,33 +424,6 @@ export function buildSidecarEnv(cfg) {
     if (cfg.compactModel && !env.LIBRAVDB_SUMMARIZER_MODEL) {
         env.LIBRAVDB_SUMMARIZER_MODEL = cfg.compactModel;
     }
-    if (cfg.gatingWeights?.w1c != null) {
-        env.LIBRAVDB_GATING_W1C = String(cfg.gatingWeights.w1c);
-    }
-    if (cfg.gatingWeights?.w2c != null) {
-        env.LIBRAVDB_GATING_W2C = String(cfg.gatingWeights.w2c);
-    }
-    if (cfg.gatingWeights?.w3c != null) {
-        env.LIBRAVDB_GATING_W3C = String(cfg.gatingWeights.w3c);
-    }
-    if (cfg.gatingWeights?.w1t != null) {
-        env.LIBRAVDB_GATING_W1T = String(cfg.gatingWeights.w1t);
-    }
-    if (cfg.gatingWeights?.w2t != null) {
-        env.LIBRAVDB_GATING_W2T = String(cfg.gatingWeights.w2t);
-    }
-    if (cfg.gatingWeights?.w3t != null) {
-        env.LIBRAVDB_GATING_W3T = String(cfg.gatingWeights.w3t);
-    }
-    if (typeof cfg.gatingTechNorm === "number" && cfg.gatingTechNorm > 0) {
-        env.LIBRAVDB_GATING_TECH_NORM = String(cfg.gatingTechNorm);
-    }
-    if (typeof cfg.ingestionGateThreshold === "number" && cfg.ingestionGateThreshold >= 0) {
-        env.LIBRAVDB_GATING_THRESHOLD = String(cfg.ingestionGateThreshold);
-    }
-    if (typeof cfg.gatingCentroidK === "number" && cfg.gatingCentroidK > 0) {
-        env.LIBRAVDB_GATING_CENTROID_K = String(cfg.gatingCentroidK);
-    }
     if (typeof cfg.lifecycleJournalMaxEntries === "number" && cfg.lifecycleJournalMaxEntries > 0) {
         env.LIBRAVDB_LIFECYCLE_JOURNAL_MAX_ENTRIES = String(cfg.lifecycleJournalMaxEntries);
     }

package/dist/types.d.ts

@@ -52,16 +52,6 @@ export interface PluginConfig {
     dreamPromotionDiaryPath?: string;
     dreamPromotionUserId?: string;
     dreamPromotionDebounceMs?: number;
-    gatingWeights?: {
-        w1c?: number;
-        w2c?: number;
-        w3c?: number;
-        w1t?: number;
-        w2t?: number;
-        w3t?: number;
-    };
-    gatingTechNorm?: number;
-    gatingCentroidK?: number;
     lifecycleJournalMaxEntries?: number;
     compactionQualityWeight?: number;
     recencyLambdaSession?: number;

package/docs/embedding-profiles.md

@@ -1,33 +1,32 @@
 # Embedding Profiles
 
-The plugin now supports a lightweight `embeddingProfile` setting for named local model metadata defaults.
+The plugin uses `embeddingProfile` for named local model metadata defaults.
 
-Default selection baseline as of `2026-03-28`:
+Default selection baseline:
 
-- default embedding profile: `all-minilm-l6-v2`
-- bundled fallback profile: `all-minilm-l6-v2`
+- default embedding profile: `nomic-embed-text-v1.5`
+- bundled fallback profile: `bge-small-en-v1.5`
 
 Why:
 
-- MiniLM keeps the local LongMemEval retrieval slice inside the target memory envelope on macOS.
-- Nomic remains available as an explicit opt-in profile for long-context experiments.
-- Nomic ONNX on macOS is fragile with CoreML execution and can trigger multi-GB RSS, so it is no longer the safe bundled default.
+- Nomic is the default because its Matryoshka-trained embeddings deliver significantly higher retrieval accuracy than MiniLM, with principled dimensionality tiering (`64d → 256d → 768d`) that lets the daemon trade memory for precision without re-embedding.
+- bge-small-en-v1.5 is the fallback for resource-constrained systems and is automatically selected when the primary model's dimensions do not match the active collection.
 - Intel Macs without reliable Metal/MPS support should set `onnxDevice: "cpu"` to force CPU ONNX execution and bypass CoreML.
 
 Current shipped profile names:
 
-- `all-minilm-l6-v2`
-  - family: `all-minilm-l6-v2`
-  - dimensions: `384`
-  - normalize: `true`
-  - max context tokens: `128`
-
 - `nomic-embed-text-v1.5`
   - family: `nomic-embed-text-v1.5`
   - dimensions: `768`
   - normalize: `true`
   - max context tokens: `8192`
 
+- `bge-small-en-v1.5`
+  - family: `bge-small-en-v1.5`
+  - dimensions: `384`
+  - normalize: `true`
+  - max context tokens: `512`
+
 How it works:
 
 - `embeddingProfile` supplies metadata defaults like family, dimensions, and normalize behavior.
@@ -38,6 +37,6 @@ How it works:
 
 Recommended usage:
 
-- `bundled` for the shipped default path, which now prefers MiniLM for local stability.
-- `onnx-local` plus `embeddingProfile` when a power user wants a known model family like Nomic with local assets.
+- `bundled` for the shipped default path, which uses `nomic-embed-text-v1.5`.
+- `onnx-local` plus `embeddingProfile` when a power user wants a known model family with local assets.
 - treat remote/Ollama providers as future separate backend types, not as overloads of `custom-local`.

package/docs/install.md

@@ -63,7 +63,7 @@ If you run the daemon on a non-default endpoint, add a plugin config:
       "libravdb-memory": {
         "enabled": true,
         "config": {
-          "sidecarPath": "unix:/Users/<you>/.clawdb/run/libravdb.sock"
+          "sidecarPath": "unix:/Users/<you>/.libravdbd/run/libravdb.sock"
         }
       }
     }
@@ -74,10 +74,10 @@ If you run the daemon on a non-default endpoint, add a plugin config:
 When `sidecarPath` is set to `"auto"`, the plugin resolves endpoints in this order on macOS/Linux:
 
 1. `LIBRAVDB_RPC_ENDPOINT` if it is set to a valid daemon endpoint
-2. `$HOME/.clawdb/run/libravdb.sock` if it exists
-3. `/opt/homebrew/var/clawdb/run/libravdb.sock` if it exists
-4. `/usr/local/var/clawdb/run/libravdb.sock` if it exists
-5. fallback to `$HOME/.clawdb/run/libravdb.sock`
+2. `$HOME/.libravdbd/run/libravdb.sock` if it exists
+3. `/opt/homebrew/var/libravdbd/run/libravdb.sock` if it exists
+4. `/usr/local/var/libravdbd/run/libravdb.sock` if it exists
+5. fallback to `$HOME/.libravdbd/run/libravdb.sock`
 
 ## Sidecar Daemon Install
 
@@ -85,13 +85,15 @@ The daemon owns the local database, embeddings, and JSON-RPC endpoint.
 
 Default endpoints:
 
-- Homebrew on macOS: `unix:/opt/homebrew/var/clawdb/run/libravdb.sock`
-- macOS/Linux user-local installs: `unix:$HOME/.clawdb/run/libravdb.sock`
+- Homebrew on macOS (Apple Silicon): `unix:/opt/homebrew/var/libravdbd/run/libravdb.sock`
+- Homebrew on macOS (Intel): `unix:/usr/local/var/libravdbd/run/libravdb.sock`
+- macOS/Linux user-local installs: `unix:$HOME/.libravdbd/run/libravdb.sock`
 - Windows: `tcp:127.0.0.1:37421`
 
 Default data path:
 
-- macOS/Linux/Windows user installs: `$HOME/.clawdb/data.libravdb`
+- macOS/Linux user installs: `$HOME/.libravdbd/data_nomic-embed-text-v1_5.libravdb`
+- Windows user installs: `%USERPROFILE%\.libravdbd\data_nomic-embed-text-v1_5.libravdb`
 
 ### Homebrew
 
@@ -188,5 +190,5 @@ Healthy output should show that:
 
 If OpenClaw cannot reach the daemon, verify the endpoint first:
 
-- macOS/Linux default: `unix:$HOME/.clawdb/run/libravdb.sock`
+- macOS/Linux default: `unix:$HOME/.libravdbd/run/libravdb.sock`
 - Windows default: `tcp:127.0.0.1:37421`

package/docs/installation.md

@@ -87,7 +87,7 @@ If the daemon uses a non-default endpoint, add `sidecarPath`:
       "libravdb-memory": {
         "enabled": true,
         "config": {
-          "sidecarPath": "unix:/Users/<you>/.clawdb/run/libravdb.sock"
+          "sidecarPath": "unix:/Users/<you>/.libravdbd/run/libravdb.sock"
         }
       }
     }
@@ -98,23 +98,23 @@ If the daemon uses a non-default endpoint, add `sidecarPath`:
 When `sidecarPath` is `"auto"`, macOS/Linux endpoint resolution checks:
 
 1. `LIBRAVDB_RPC_ENDPOINT`
-2. `$HOME/.clawdb/run/libravdb.sock`
-3. `/opt/homebrew/var/clawdb/run/libravdb.sock`
-4. `/usr/local/var/clawdb/run/libravdb.sock`
-5. fallback to `$HOME/.clawdb/run/libravdb.sock`
+2. `$HOME/.libravdbd/run/libravdb.sock`
+3. `/opt/homebrew/var/libravdbd/run/libravdb.sock`
+4. `/usr/local/var/libravdbd/run/libravdb.sock`
+5. fallback to `$HOME/.libravdbd/run/libravdb.sock`
 
 ## Default Paths
 
 | Platform | Default endpoint |
 |---|---|
-| macOS/Linux user-local | `unix:$HOME/.clawdb/run/libravdb.sock` |
-| macOS Homebrew Apple Silicon | `unix:/opt/homebrew/var/clawdb/run/libravdb.sock` |
+| macOS/Linux user-local | `unix:$HOME/.libravdbd/run/libravdb.sock` |
+| macOS Homebrew Apple Silicon | `unix:/opt/homebrew/var/libravdbd/run/libravdb.sock` |
 | Windows | `tcp:127.0.0.1:37421` |
 
 Default data path:
 
 ```text
-$HOME/.clawdb/data.libravdb
+$HOME/.libravdbd/data_nomic-embed-text-v1_5.libravdb
 ```
 
 ## Verification
@@ -135,7 +135,7 @@ Expected output shape:
 │ Lifecycle hints    │ 0                            │
 │ Gate threshold     │ 0.35                         │
 │ Abstractive model  │ ready | not provisioned      │
-│ Embedding profile  │ all-minilm-l6-v2             │
+│ Embedding profile  │ nomic-embed-text-v1.5             │
 │ Message            │ ok                           │
 └────────────────────┴──────────────────────────────┘
 ```

package/docs/models.md

@@ -20,16 +20,15 @@ latency and offline operation are part of the product contract.
 
 ## Default And Optional Embedding Profiles
 
-The current safe default profile is `all-minilm-l6-v2`.
+The default profile is `nomic-embed-text-v1.5`. Nomic was chosen as the default
+because its Matryoshka-trained embeddings deliver significantly higher retrieval
+accuracy than MiniLM, with principled dimensionality tiering (`64d → 256d →
+768d`) that lets the daemon trade memory for precision without re-embedding.
 
-MiniLM is the default because it keeps local retrieval within the target memory
-envelope on macOS and is less fragile with ONNX Runtime execution than larger
-profiles.
-
-`nomic-embed-text-v1.5` remains available as an explicit opt-in profile for
-long-context retrieval experiments. Nomic's Matryoshka training makes
-`64d -> 256d -> 768d` tiering principled rather than arbitrary truncation, but
-its larger footprint makes it a less conservative default.
+`bge-small-en-v1.5` is the fallback profile. It has a smaller disk and memory
+footprint than Nomic and is automatically selected when the primary model's
+dimensions do not match the active collection. Operators on resource-constrained
+systems can also set it as the primary profile for lighter local inference.
 
 For exact profile metadata, read [Embedding profiles](./embedding-profiles.md).
 
@@ -46,8 +45,8 @@ generative models would increase latency and operational complexity.
 
 | Model/profile | Role |
 |---|---|
-| `all-minilm-l6-v2` | Default lightweight embedding profile. |
-| `nomic-embed-text-v1.5` | Opt-in long-context embedding profile. |
+| `nomic-embed-text-v1.5` | Default embedding profile — high-accuracy Matryoshka embeddings. |
+| `bge-small-en-v1.5` | Fallback embedding profile — lighter footprint for constrained systems. |
 | T5-small | Optional local abstractive compaction summarizer. |
 
 External summarizer endpoints, such as Ollama, are optional. They are not part

package/docs/security.md

@@ -27,7 +27,7 @@ Current implementation facts:
 - the published plugin manifest does not register `openclaw.setup`
 - the published plugin source contains no direct `child_process` usage
 - the plugin connects only to a configured local endpoint such as
-  `unix:/Users/<you>/.clawdb/run/libravdb.sock` or `tcp:127.0.0.1:37421`
+  `unix:/Users/<you>/.libravdbd/run/libravdb.sock` or `tcp:127.0.0.1:37421`
 - daemon installation and lifecycle are explicit user or operator actions
 
 The daemon distribution surface should be evaluated separately from the plugin

package/docs/uninstall.md

@@ -84,8 +84,8 @@ Only do this if you want to permanently remove stored LibraVDB memory.
 
 Common local state:
 
-- socket directory: `~/.clawdb/run/`
-- database file: `~/.clawdb/data.libravdb`
+- socket directory: `~/.libravdbd/run/`
+- database file: `~/.libravdbd/data_nomic-embed-text-v1_5.libravdb`
 
 If you configured a custom Unix socket endpoint in `sidecarPath`, remove that
 socket path or containing directory if applicable. If you configured `dbPath`,

package/openclaw.plugin.json

@@ -2,7 +2,7 @@
   "id": "libravdb-memory",
   "name": "LibraVDB Memory",
   "description": "Persistent vector memory with three-tier hybrid scoring",
-  "version": "1.4.30",
+  "version": "1.4.32",
   "kind": [
     "memory",
     "context-engine"
@@ -85,11 +85,11 @@
       },
       "embeddingProfile": {
         "type": "string",
-        "default": "all-minilm-l6-v2"
+        "default": "nomic-embed-text-v1.5"
       },
       "fallbackProfile": {
         "type": "string",
-        "default": "all-minilm-l6-v2"
+        "default": "bge-small-en-v1.5"
       },
       "embeddingModelPath": {
         "type": "string"
@@ -215,52 +215,6 @@
         "type": "number",
         "default": 150
       },
-      "gatingWeights": {
-        "type": "object",
-        "additionalProperties": false,
-        "default": {
-          "w1c": 0.35,
-          "w2c": 0.4,
-          "w3c": 0.25,
-          "w1t": 0.4,
-          "w2t": 0.35,
-          "w3t": 0.25
-        },
-        "properties": {
-          "w1c": {
-            "type": "number",
-            "default": 0.35
-          },
-          "w2c": {
-            "type": "number",
-            "default": 0.4
-          },
-          "w3c": {
-            "type": "number",
-            "default": 0.25
-          },
-          "w1t": {
-            "type": "number",
-            "default": 0.4
-          },
-          "w2t": {
-            "type": "number",
-            "default": 0.35
-          },
-          "w3t": {
-            "type": "number",
-            "default": 0.25
-          }
-        }
-      },
-      "gatingTechNorm": {
-        "type": "number",
-        "default": 1.5
-      },
-      "gatingCentroidK": {
-        "type": "number",
-        "default": 10
-      },
       "lifecycleJournalMaxEntries": {
         "type": "number",
         "default": 500

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xdarkicex/openclaw-memory-libravdb",
-  "version": "1.4.30",
+  "version": "1.4.32",
   "type": "module",
   "main": "./dist/index.js",
   "types": "./dist/index.d.ts",