mcp-super-memory 0.4.8 → 0.5.1

package/README.md

@@ -1,7 +1,7 @@
 # mcp-super-memory
 
-[![PyPI version](https://img.shields.io/pypi/v/mcp-super-memory)](https://pypi.org/project/mcp-super-memory/)
-[![Python](https://img.shields.io/pypi/pyversions/mcp-super-memory)](https://pypi.org/project/mcp-super-memory/)
+[![npm version](https://img.shields.io/npm/v/mcp-super-memory)](https://www.npmjs.com/package/mcp-super-memory)
+[![Node.js](https://img.shields.io/badge/Node.js-20%2B-green)](https://nodejs.org/)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
 
 **N:M associative memory graph for LLM agents — delivered as an MCP server.**
@@ -63,7 +63,9 @@ Search `"Newton"` → matches `[Newton]`, `[apple]` keys (1-hop) → follows sha
 | Time decay | ✅ depth-weighted | ❌ | ❌ | ❌ |
 | Key types | ✅ concept/name/proper_noun | ❌ | ❌ | ❌ |
 | Key merge (IDF) | ✅ | ❌ | ❌ | ❌ |
-| Dual-path recall | ✅ key + content | ❌ | ❌ | ❌ |
+| Hybrid retrieval (BM25 + dense + RRF) | ✅ | ❌ | partial | ❌ |
+| Dual-path dense recall | ✅ key + content | ❌ | ❌ | ❌ |
+| Hebbian link learning | ✅ | ❌ | ❌ | ❌ |
 
 ### Depth System
 
@@ -83,7 +85,7 @@ Not all keys should behave the same. Names shouldn't match semantically — "동
 
 | Type | Matching | Use Case |
 |------|----------|----------|
-| `concept` (default) | Embedding similarity ≥ 0.35 | Topics, categories, attributes |
+| `concept` (default) | Embedding similarity ≥ threshold (0.28 OpenAI / 0.60 local) | Topics, categories, attributes |
 | `name` | Exact match only | Person names |
 | `proper_noun` | Exact match only | Brands, places |
 
@@ -108,14 +110,26 @@ Add key "파이썬"  → finds existing "Python" (similarity 0.87 > threshold 0.
 
 Prevents key space fragmentation. Same concept across languages or phrasing stays unified.
 
-### Dual-Path Recall
+### Hybrid Retrieval (BM25 + dense, RRF-fused)
 
-Recall searches two paths simultaneously:
+Recall is not a single similarity scan. Three signals run in parallel and are fused with **Reciprocal Rank Fusion** (`RRF_K = 60`):
 
-- **Path A (key matching):** Query embedding → match keys → follow links → memories
-- **Path B (content matching):** Query embedding → directly compare against memory content embeddings
+- **BM25 (sparse):** lexical full-text search over memory content (MiniSearch, fuzzy + prefix). Catches exact terms, names, and rare tokens that embeddings blur.
+- **Dense Path A (key matching):** query embedding → match keys → follow links → memories. Score = `keySim × IDF × linkWeight`, summed across all matching keys.
+- **Dense Path B (content matching):** query embedding → directly compare against memory content embeddings. Finds memories even when they weren't tagged with the right keys.
 
-Scores from both paths are summed. This ensures memories are found even when they weren't tagged with the right keys.
+Sparse and dense rank lists are merged by RRF, then modulated by depth and time before 2-hop expansion. Combining lexical and semantic signals is more robust than either alone.
+
+### Hebbian Link Learning
+
+Recall is a **write**, not just a read. Every recall reshapes the graph:
+
+- Links whose key **actually matched the query** and led to a **returned** memory are **reinforced** (`+0.1`, capped at `3.0`).
+- Links explored (matched key) but whose memory was **not returned** are **decayed** (`−0.005`, floored at `0.1`).
+
+Reinforcement is scoped to the keys that *fired* for this query — not to every key of a returned memory. This is the literal Hebbian rule ("fire together, wire together") and it matters: reinforcing a returned memory's unrelated keys would let a stray association grow every time that memory surfaced for a *different* key, slowly polluting the graph. Weights are clamped to `[0.1, 3.0]`, so a hot memory's pull is bounded and the graph can recover from a bad reinforcement via subsequent decay.
+
+Link weights feed directly back into scoring (`keySim × IDF × linkWeight`), so connections that repeatedly co-fire grow stronger and stale ones fade — the graph learns which associations actually matter from access patterns.
 
 ---
 
@@ -138,29 +152,62 @@ Scores from both paths are summed. This ensures memories are found even when the
 └─────────────────────────────────────────────────────────┘
 ```
 
-**Recall algorithm (2-hop):**
-1. Embed query → find matching keys (concept: similarity ≥ 0.35, name/proper_noun: exact match)
-2. Also compare query embedding directly against memory content embeddings (≥ 0.3)
-3. Follow links → collect memories, aggregate scores (multiple key matches sum up, IDF-weighted)
-4. For each 1-hop memory: follow *its* keys → find 2-hop memories (score × `HOP_DECAY = 0.3`)
-5. Apply depth factor (`0.5 + depth × 0.5`) and time decay (depth-weighted, 30-day half-life)
-6. Return ranked results with `hop` field
+**Recall algorithm (hybrid, 2-hop):**
+
+Three retrieval signals run in parallel, then get fused and expanded:
+
+1. **BM25 (sparse):** lexical search over memory content (MiniSearch, fuzzy `0.2` + prefix). Keep top 50.
+2. **Dense Path A (keys):** embed query → match keys (concept: cosine ≥ threshold; name/proper_noun: substring match → score `1.0`) → take top 10 keys → follow links. Score = `keySim × IDF × linkWeight`, summed across matching keys.
+3. **Dense Path B (content):** compare query embedding directly against memory content embeddings (cosine ≥ threshold).
+4. **RRF fusion:** merge the BM25 and dense rank lists via `score += 1 / (RRF_K + rank + 1)` (`RRF_K = 60`).
+5. **Depth & time modulation:** `score × (0.9 + depth × 0.1) × timeFactor`, where `timeFactor` is a depth-weighted 30-day half-life decay (deep memories decay slower).
+6. **2-hop expansion:** for each fused memory, follow *its* keys to associated memories (`× HOP_DECAY(0.3) × IDF × linkWeight`) and traverse explicit `related_to` links (bidirectional, `× HOP_DECAY`).
+7. **Hebbian update:** reinforce matched-key links of returned memories (`+0.1`), decay explored-but-unreturned links (`−0.005`).
+8. Return ranked results with `hop` field (`1` = direct, `2` = associative).
+
+### Similarity thresholds (calibrated per embedding model)
+
+Embedding backends have very different cosine distributions, so a single threshold set cannot serve all of them. The thresholds below are calibrated per model (`getThresholdProfile()` in `src/embedding.ts`):
+
+| Threshold | OpenAI | Local BGE (en) | Local e5 (multilingual) |
+|-----------|--------|----------------|--------------------------|
+| Key recall (query↔key cosine) | 0.28 | 0.60 | 0.85 |
+| Content recall (query↔content cosine) | 0.28 | 0.50 | 0.80 |
+| Key auto-link | 0.50 | 0.60 | 0.93 |
+| Key merge | 0.85 | 0.85 | 0.97 |
+| Memory dedup | 0.90 | 0.90 | 0.985 |
+
+**Why e5 differs so much:** multilingual-e5 packs embeddings into a narrow high-cosine band (~0.86–0.99). Same-word query↔key pairs (asymmetric `query:`/`passage:` prefixes) still separate cleanly (~0.89 vs ≤0.82), but key↔key and content↔content do **not** — distinct facts like *"A uses Postgres"* and *"B uses Mongo"* sit at ~0.96, dangerously close to true paraphrases (~0.99). Hence e5's merge/dedup/auto-link thresholds are pushed high to avoid silently collapsing distinct memories.
+
+**Drift escape hatch:** if you switch models or your data's character drifts, override any threshold without code changes:
+
+```
+SUPER_MEMORY_KEY_RECALL=0.82
+SUPER_MEMORY_MEMORY_DEDUP=0.99
+# also: _KEY_MERGE, _KEY_AUTOLINK, _CONTENT_RECALL  (values in [0,1])
+```
+
+An uncalibrated `LOCAL_EMBEDDING_MODEL` falls back to the BGE profile **and logs a warning** so the miscalibration is never silent.
+
+> **Multilingual note:** cross-lingual *content* matching has a same-language bias (a Korean query scores Korean memories higher regardless of meaning). The reliable cross-lingual path is the **key graph** — tag memories with keys in multiple languages (e.g. `["딸기", "strawberry"]`) so recall hits the key exactly instead of relying on biased content similarity.
 
 ---
 
 ## MCP Tools
 
-The memory system exposes 8 tools via MCP:
+The memory system exposes 10 tools via MCP:
 
 | Tool | Description |
-|------|-------------|
-| `recall(query, top_k)` | N:M search with 2-hop associative traversal + content matching |
-| `remember(content, keys, key_types?)` | Save memory with key concepts and optional type annotations |
-| `correct(memory_id, content, keys?)` | Versioned update — old memory preserved but weakened |
+| --- | --- |
+| `recall(query, top_k, namespace?, expand?)` | Hybrid search (BM25 + dense key/content, RRF-fused) with 2-hop associative traversal |
+| `remember(content, keys, key_types?, namespace?, ttl_seconds?, related_to?)` | Save memory with key concepts and optional type annotations |
+| `correct(memory_id, content, keys?, key_types?, related_to?)` | Versioned update — old memory preserved but weakened |
 | `related(memory_id)` | Find memories sharing keys (associative exploration) |
 | `forget(memory_id)` | Permanently delete |
 | `get_conversation(session_id, turn?)` | Load original conversation turns |
-| `list_memories()` | List all stored memories with keys, depth, access count |
+| `list_memories(namespace?)` | List all stored memories with keys, depth, access count |
+| `remember_batch(items)` | Save multiple memories in one call |
+| `cleanup_expired()` | Delete memories whose TTL has expired |
 | `memory_stats()` | Get current key/memory/link counts |
 
 A system prompt template is also available via `memory_system_prompt` MCP prompt — include it to instruct the agent to recall silently, use diverse keys, and never mention the memory system to users.
@@ -178,8 +225,8 @@ Add to `claude_desktop_config.json`:
 {
   "mcpServers": {
     "mcp-super-memory": {
-      "command": "uvx",
-      "args": ["mcp-super-memory"],
+      "command": "npx",
+      "args": ["-y", "mcp-super-memory"],
       "env": {
         "OPENAI_API_KEY": "your-openai-api-key"
       }
@@ -193,8 +240,8 @@ Add to `claude_desktop_config.json`:
 {
   "mcpServers": {
     "mcp-super-memory": {
-      "command": "uvx",
-      "args": ["mcp-super-memory[local]"],
+      "command": "npx",
+      "args": ["-y", "mcp-super-memory"],
       "env": {
         "EMBEDDING_BACKEND": "local"
       }
@@ -207,10 +254,10 @@ Add to `claude_desktop_config.json`:
 
 ```bash
 # OpenAI embeddings
-claude mcp add mcp-super-memory -e OPENAI_API_KEY=your-openai-api-key -- uvx mcp-super-memory
+claude mcp add mcp-super-memory -e OPENAI_API_KEY=your-openai-api-key -- npx -y mcp-super-memory
 
 # Local embeddings (no API key required)
-claude mcp add mcp-super-memory -e EMBEDDING_BACKEND=local -- uvx "mcp-super-memory[local]"
+claude mcp add mcp-super-memory -e EMBEDDING_BACKEND=local -- npx -y mcp-super-memory
 ```
 
 ### Manual / Development
@@ -218,6 +265,7 @@ claude mcp add mcp-super-memory -e EMBEDDING_BACKEND=local -- uvx "mcp-super-mem
 ```bash
 git clone https://github.com/donggyun112/mcp-super-memory
 cd super-memory
+pnpm install
 ```
 
 Create `.env`:
@@ -229,19 +277,25 @@ OPENAI_EMBEDDING_MODEL=text-embedding-3-small
 Or use local embeddings (no API key required):
 ```
 EMBEDDING_BACKEND=local
-LOCAL_EMBEDDING_MODEL=paraphrase-multilingual-MiniLM-L12-v2  # optional, this is the default
+LOCAL_EMBEDDING_MODEL=fast-multilingual-e5-large  # default; best fit for Korean/multilingual keys
 ```
 
-> **Note:** Mixing backends on existing data will break recall. If switching backends, clear `~/.super-memory/graph.json` first.
+If `OPENAI_API_KEY` is not set and `EMBEDDING_BACKEND` is unset, the server automatically uses the local `fastembed` backend.
+For English-only use or lower local resource usage, set `LOCAL_EMBEDDING_MODEL=fast-bge-base-en-v1.5` or `fast-bge-small-en-v1.5`.
+
+> **Switching backends is safe.** If you change the embedding backend/model, on next startup the graph **auto-migrates** — every key and memory is re-embedded with the new backend while content, links, depth, and access history are preserved (a `graph.json.bak.<dim>d` backup is written first). Disable with `SUPER_MEMORY_AUTO_MIGRATE=false`. Re-embedding via OpenAI incurs one-time API cost proportional to your memory count.
 
 ```bash
-uv sync
-uv run mcp-super-memory
+pnpm dev
+# or:
+pnpm build
+pnpm start
 ```
 
 **Requirements:**
-- Python 3.12+
-- OpenAI API key (for embeddings) — or `sentence-transformers` for local embeddings
+- Node.js 20+
+- pnpm for local development
+- OpenAI API key for OpenAI embeddings, or `fastembed` for local embeddings
 
 ---
 
@@ -250,12 +304,14 @@ uv run mcp-super-memory
 All data is local. No external database required.
 
 ```
-data/
+~/.super-memory/
 ├── graph.json          # keys, memories, links
 └── conversations/
     └── {session_id}.jsonl   # original conversation turns
 ```
 
+Set `SUPER_MEMORY_DATA_DIR` to use a different storage directory.
+
 ---
 
 ## Limitations
@@ -263,6 +319,22 @@ data/
 - **Linear scan** — suitable for personal use (~10k memories). FAISS/ChromaDB integration planned for larger scale.
 - **2-hop max** — deeper associative chains require `related()` tool calls by the agent.
 - **Agent quality matters** — key selection on `remember` affects retrieval quality. System prompt tuning is important.
+- **Cross-lingual content bias** — with multilingual e5, raw content similarity favors same-language memories regardless of meaning. Tag memories with multilingual keys so the key graph (not biased content cosine) carries cross-lingual recall.
+- **Threshold calibration** — thresholds are tuned per embedding model. A new/uncalibrated model falls back to the BGE profile (with a warning); recalibrate via the `SUPER_MEMORY_*` env overrides.
+
+---
+
+## Testing
+
+```bash
+pnpm test                        # unit tests (fast, no model download)
+tsx test/scenarios.ts            # 21 end-to-end behavioral checks (local e5)
+tsx test/robustness.ts           # threshold overrides + Hebbian pollution bounds
+tsx test/migration.ts            # backend/dimension switch auto-migration (no brick)
+tsx test/live-multilingual.ts    # interactive multilingual recall demo
+```
+
+`scenarios.ts` and `robustness.ts` exercise the real local embedding backend (direct/associative/cross-lingual recall, versioning, depth growth, dedup, TTL, Hebbian learning, namespace isolation). They double as a recalibration harness when tuning thresholds for a new model.
 
 ---
 

package/dist/embedding.d.ts

@@ -1,5 +1,17 @@
 export declare const OPENAI_API_KEY: string;
 export declare const OPENAI_EMBEDDING_MODEL: string;
 export declare const EMBEDDING_BACKEND: string;
-export declare function embedTextAsync(text: string): Promise<number[]>;
+export declare const LOCAL_EMBEDDING_MODEL: string;
+export declare const LOCAL_EMBEDDING_CACHE_DIR: string;
+type EmbeddingInputType = "passage" | "query";
+export interface ThresholdProfile {
+    keyMerge: number;
+    memoryDedup: number;
+    keyAutoLink: number;
+    keyRecall: number;
+    contentRecall: number;
+}
+export declare function getThresholdProfile(): ThresholdProfile;
+export declare function embedTextAsync(text: string, inputType?: EmbeddingInputType): Promise<number[]>;
+export {};
 //# sourceMappingURL=embedding.d.ts.map

package/dist/embedding.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"embedding.d.ts","sourceRoot":"","sources":["../src/embedding.ts"],"names":[],"mappings":"AAKA,eAAO,MAAM,cAAc,QAAmC,CAAC;AAC/D,eAAO,MAAM,sBAAsB,QAC6B,CAAC;AAKjE,eAAO,MAAM,iBAAiB,QACqB,CAAC;AAsCpD,wBAAsB,cAAc,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,CAqBpE"}
+{"version":3,"file":"embedding.d.ts","sourceRoot":"","sources":["../src/embedding.ts"],"names":[],"mappings":"AAKA,eAAO,MAAM,cAAc,QAAmC,CAAC;AAC/D,eAAO,MAAM,sBAAsB,QAC6B,CAAC;AAGjE,eAAO,MAAM,iBAAiB,QACqB,CAAC;AACpD,eAAO,MAAM,qBAAqB,QACiC,CAAC;AACpE,eAAO,MAAM,yBAAyB,QACkB,CAAC;AAGzD,KAAK,kBAAkB,GAAG,SAAS,GAAG,OAAO,CAAC;AAmE9C,MAAM,WAAW,gBAAgB;IAC/B,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,MAAM,CAAC;IACpB,SAAS,EAAE,MAAM,CAAC;IAClB,aAAa,EAAE,MAAM,CAAC;CACvB;AAkDD,wBAAgB,mBAAmB,IAAI,gBAAgB,CAYtD;AA6CD,wBAAsB,cAAc,CAClC,IAAI,EAAE,MAAM,EACZ,SAAS,GAAE,kBAA8B,GACxC,OAAO,CAAC,MAAM,EAAE,CAAC,CAqBnB"}

package/dist/embedding.js

@@ -3,10 +3,25 @@ config();
 import OpenAI from "openai";
 export const OPENAI_API_KEY = process.env.OPENAI_API_KEY ?? "";
 export const OPENAI_EMBEDDING_MODEL = process.env.OPENAI_EMBEDDING_MODEL ?? "text-embedding-3-small";
-const LOCAL_EMBEDDING_MODEL = process.env.LOCAL_EMBEDDING_MODEL ?? "paraphrase-multilingual-MiniLM-L12-v2";
 const _DEFAULT_BACKEND = OPENAI_API_KEY ? "openai" : "local";
 export const EMBEDDING_BACKEND = process.env.EMBEDDING_BACKEND ?? _DEFAULT_BACKEND;
+export const LOCAL_EMBEDDING_MODEL = process.env.LOCAL_EMBEDDING_MODEL ?? "fast-multilingual-e5-large";
+export const LOCAL_EMBEDDING_CACHE_DIR = process.env.LOCAL_EMBEDDING_CACHE_DIR ?? "local_cache";
 const EMBED_RETRIES = 3;
+const LOCAL_MODEL_ALIASES = {
+    "fast-multilingual-e5-large": "MLE5Large",
+    "multilingual-e5-large": "MLE5Large",
+    mle5large: "MLE5Large",
+    "fast-bge-base-en-v1.5": "BGEBaseENV15",
+    "bge-base-en-v1.5": "BGEBaseENV15",
+    bgebaseenv15: "BGEBaseENV15",
+    "fast-bge-small-en-v1.5": "BGESmallENV15",
+    "bge-small-en-v1.5": "BGESmallENV15",
+    bgesmallenv15: "BGESmallENV15",
+    "fast-all-minilm-l6-v2": "AllMiniLML6V2",
+    "all-minilm-l6-v2": "AllMiniLML6V2",
+    allminilml6v2: "AllMiniLML6V2",
+};
 let _openaiClient = null;
 function getOpenAIClient() {
     if (!_openaiClient) {
@@ -14,30 +29,128 @@ function getOpenAIClient() {
     }
     return _openaiClient;
 }
+function errorMessage(err) {
+    return err instanceof Error ? err.message : String(err);
+}
+function normalizeModelName(modelName) {
+    return modelName
+        .trim()
+        .replace(/^EmbeddingModel\./i, "")
+        .toLowerCase()
+        .replace(/[\s_]/g, "-");
+}
+function resolveLocalModel(modelName, embeddingModel) {
+    const normalized = normalizeModelName(modelName);
+    const aliasKey = LOCAL_MODEL_ALIASES[normalized];
+    if (aliasKey && aliasKey in embeddingModel)
+        return embeddingModel[aliasKey];
+    for (const [key, value] of Object.entries(embeddingModel)) {
+        if (key.toLowerCase() === modelName.trim().toLowerCase())
+            return value;
+        if (value.toLowerCase() === normalized)
+            return value;
+    }
+    const supported = Object.values(embeddingModel)
+        .filter((value) => value !== embeddingModel.CUSTOM)
+        .join(", ");
+    throw new Error(`Unsupported LOCAL_EMBEDDING_MODEL "${modelName}". Supported values: ${supported}`);
+}
+const THRESHOLD_PROFILES = {
+    openai: { keyMerge: 0.85, memoryDedup: 0.9, keyAutoLink: 0.5, keyRecall: 0.28, contentRecall: 0.28 },
+    bge: { keyMerge: 0.85, memoryDedup: 0.9, keyAutoLink: 0.6, keyRecall: 0.6, contentRecall: 0.5 },
+    // e5: query↔key is asymmetric (query/passage prefixes) and separates well —
+    // exact-term matches land ~0.886+, distinct words ≤0.82, so keyRecall 0.85
+    // catches real key hits. But key↔key and content↔key are both passage-embedded
+    // and do NOT separate, so keyMerge/keyAutoLink stay high to prevent collapse.
+    // memoryDedup 0.985: e5 packs distinct-but-similar facts ("A uses Postgres" vs
+    // "B uses Mongo" ≈0.96) dangerously close to true paraphrases (≈0.99). Dedup
+    // wrongly below 0.985 would silently supersede distinct memories → data loss.
+    e5: { keyMerge: 0.97, memoryDedup: 0.985, keyAutoLink: 0.93, keyRecall: 0.85, contentRecall: 0.8 },
+    minilm: { keyMerge: 0.85, memoryDedup: 0.9, keyAutoLink: 0.6, keyRecall: 0.5, contentRecall: 0.45 },
+};
+let _warnedUncalibrated = false;
+function localModelFamily() {
+    const alias = LOCAL_MODEL_ALIASES[normalizeModelName(LOCAL_EMBEDDING_MODEL)];
+    if (alias === "MLE5Large")
+        return "e5";
+    if (alias === "AllMiniLML6V2")
+        return "minilm";
+    if (alias === "BGEBaseENV15" || alias === "BGESmallENV15")
+        return "bge";
+    // Unknown model: thresholds are NOT calibrated for it. A wrong profile can
+    // silently collapse (over-merge) or fragment (under-recall) the graph, so make
+    // the miscalibration loud and point at the env override escape hatch.
+    if (!_warnedUncalibrated) {
+        _warnedUncalibrated = true;
+        console.error(`[super-memory] WARNING: no calibrated threshold profile for ` +
+            `LOCAL_EMBEDDING_MODEL="${LOCAL_EMBEDDING_MODEL}". Falling back to BGE ` +
+            `thresholds — the graph may mis-cluster. Override per-threshold with ` +
+            `SUPER_MEMORY_KEY_MERGE / _MEMORY_DEDUP / _KEY_AUTOLINK / _KEY_RECALL / _CONTENT_RECALL.`);
+    }
+    return "bge";
+}
+function envThreshold(name) {
+    const raw = process.env[name];
+    if (raw === undefined || raw.trim() === "")
+        return undefined;
+    const n = Number(raw);
+    if (!Number.isFinite(n) || n < 0 || n > 1) {
+        console.error(`[super-memory] WARNING: ignoring ${name}="${raw}" (must be a number in [0,1]).`);
+        return undefined;
+    }
+    return n;
+}
+// Per-threshold env overrides are an escape hatch for when the model or the
+// character of stored data drifts away from what the built-in profiles assume.
+export function getThresholdProfile() {
+    const base = EMBEDDING_BACKEND !== "local"
+        ? THRESHOLD_PROFILES.openai
+        : THRESHOLD_PROFILES[localModelFamily()];
+    return {
+        keyMerge: envThreshold("SUPER_MEMORY_KEY_MERGE") ?? base.keyMerge,
+        memoryDedup: envThreshold("SUPER_MEMORY_MEMORY_DEDUP") ?? base.memoryDedup,
+        keyAutoLink: envThreshold("SUPER_MEMORY_KEY_AUTOLINK") ?? base.keyAutoLink,
+        keyRecall: envThreshold("SUPER_MEMORY_KEY_RECALL") ?? base.keyRecall,
+        contentRecall: envThreshold("SUPER_MEMORY_CONTENT_RECALL") ?? base.contentRecall,
+    };
+}
 // eslint-disable-next-line @typescript-eslint/no-explicit-any
 let _localModel = null;
 async function getLocalModel() {
     if (!_localModel) {
         try {
-            const { pipeline } = await import("@xenova/transformers");
-            _localModel = await pipeline("feature-extraction", LOCAL_EMBEDDING_MODEL);
+            const { FlagEmbedding, EmbeddingModel } = await import("fastembed");
+            const model = resolveLocalModel(LOCAL_EMBEDDING_MODEL, EmbeddingModel);
+            _localModel = await FlagEmbedding.init({
+                model: model,
+                cacheDir: LOCAL_EMBEDDING_CACHE_DIR,
+            });
         }
-        catch {
-            throw new Error("@xenova/transformers is not installed.\n" +
-                "Install it with: npm install @xenova/transformers\n" +
-                "Or set OPENAI_API_KEY to use OpenAI embeddings.");
+        catch (err) {
+            throw new Error("Failed to initialize local fastembed model.\n" +
+                "Install with: npm install fastembed\n" +
+                "Or set OPENAI_API_KEY to use OpenAI embeddings.\n" +
+                `Cause: ${errorMessage(err)}`);
         }
     }
     return _localModel;
 }
-async function embedLocal(text) {
+async function embedLocal(text, inputType) {
     const model = await getLocalModel();
-    const output = await model(text, { pooling: "mean", normalize: true });
-    return Array.from(output.data[0]);
+    if (inputType === "query" && typeof model.queryEmbed === "function") {
+        return Array.from(await model.queryEmbed(text));
+    }
+    const gen = typeof model.passageEmbed === "function"
+        ? model.passageEmbed([text], 256)
+        : model.embed([text]);
+    for await (const batch of gen) {
+        return Array.from(batch[0]);
+    }
+    throw new Error("fastembed returned no embeddings");
 }
-export async function embedTextAsync(text) {
+export async function embedTextAsync(text, inputType = "passage") {
     if (EMBEDDING_BACKEND === "local") {
-        return embedLocal(text);
+        return embedLocal(text, inputType);
     }
     const client = getOpenAIClient();
     for (let attempt = 0; attempt < EMBED_RETRIES; attempt++) {

package/dist/embedding.js.map

@@ -1 +1 @@
-{"version":3,"file":"embedding.js","sourceRoot":"","sources":["../src/embedding.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,MAAM,EAAE,MAAM,QAAQ,CAAC;AAChC,MAAM,EAAE,CAAC;AAET,OAAO,MAAM,MAAM,QAAQ,CAAC;AAE5B,MAAM,CAAC,MAAM,cAAc,GAAG,OAAO,CAAC,GAAG,CAAC,cAAc,IAAI,EAAE,CAAC;AAC/D,MAAM,CAAC,MAAM,sBAAsB,GACjC,OAAO,CAAC,GAAG,CAAC,sBAAsB,IAAI,wBAAwB,CAAC;AACjE,MAAM,qBAAqB,GACzB,OAAO,CAAC,GAAG,CAAC,qBAAqB,IAAI,uCAAuC,CAAC;AAE/E,MAAM,gBAAgB,GAAG,cAAc,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,OAAO,CAAC;AAC7D,MAAM,CAAC,MAAM,iBAAiB,GAC5B,OAAO,CAAC,GAAG,CAAC,iBAAiB,IAAI,gBAAgB,CAAC;AAEpD,MAAM,aAAa,GAAG,CAAC,CAAC;AAExB,IAAI,aAAa,GAAkB,IAAI,CAAC;AAExC,SAAS,eAAe;IACtB,IAAI,CAAC,aAAa,EAAE,CAAC;QACnB,aAAa,GAAG,IAAI,MAAM,CAAC,EAAE,MAAM,EAAE,cAAc,EAAE,CAAC,CAAC;IACzD,CAAC;IACD,OAAO,aAAa,CAAC;AACvB,CAAC;AAED,8DAA8D;AAC9D,IAAI,WAAW,GAAQ,IAAI,CAAC;AAE5B,KAAK,UAAU,aAAa;IAC1B,IAAI,CAAC,WAAW,EAAE,CAAC;QACjB,IAAI,CAAC;YACH,MAAM,EAAE,QAAQ,EAAE,GAAG,MAAM,MAAM,CAAC,sBAAsB,CAAC,CAAC;YAC1D,WAAW,GAAG,MAAM,QAAQ,CAAC,oBAAoB,EAAE,qBAAqB,CAAC,CAAC;QAC5E,CAAC;QAAC,MAAM,CAAC;YACP,MAAM,IAAI,KAAK,CACb,0CAA0C;gBACxC,qDAAqD;gBACrD,iDAAiD,CACpD,CAAC;QACJ,CAAC;IACH,CAAC;IACD,OAAO,WAAW,CAAC;AACrB,CAAC;AAED,KAAK,UAAU,UAAU,CAAC,IAAY;IACpC,MAAM,KAAK,GAAG,MAAM,aAAa,EAAE,CAAC;IACpC,MAAM,MAAM,GAAG,MAAM,KAAK,CAAC,IAAI,EAAE,EAAE,OAAO,EAAE,MAAM,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;IACvE,OAAO,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,CAAa,CAAC;AAChD,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,cAAc,CAAC,IAAY;IAC/C,IAAI,iBAAiB,KAAK,OAAO,EAAE,CAAC;QAClC,OAAO,UAAU,CAAC,IAAI,CAAC,CAAC;IAC1B,CAAC;IAED,MAAM,MAAM,GAAG,eAAe,EAAE,CAAC;IACjC,KAAK,IAAI,OAAO,GAAG,CAAC,EAAE,OAAO,GAAG,aAAa,EAAE,OAAO,EAAE,EAAE,CAAC;QACzD,IAAI,CAAC;YACH,MAAM,IAAI,GAAG,MAAM,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC;gBAC1C,KAAK,EAAE,sBAAsB;gBAC7B,KAAK,EAAE,IAAI;aACZ,CAAC,CAAC;YACH,OAAO,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;QAChC,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,OAAO,KAAK,aAAa,GAAG,CAAC;gBAAE,MAAM,GAAG,CAAC;YAC7C,MAAM,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,EAAE,CAC5B,UAAU,CAAC,OAAO,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,CAAC,GAAG,IAAI,CAAC,CACjD,CAAC;QACJ,CAAC;IACH,CAAC;IACD,MAAM,IAAI,KAAK,CAAC,aAAa,CAAC,CAAC;AACjC,CAAC"}
+{"version":3,"file":"embedding.js","sourceRoot":"","sources":["../src/embedding.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,MAAM,EAAE,MAAM,QAAQ,CAAC;AAChC,MAAM,EAAE,CAAC;AAET,OAAO,MAAM,MAAM,QAAQ,CAAC;AAE5B,MAAM,CAAC,MAAM,cAAc,GAAG,OAAO,CAAC,GAAG,CAAC,cAAc,IAAI,EAAE,CAAC;AAC/D,MAAM,CAAC,MAAM,sBAAsB,GACjC,OAAO,CAAC,GAAG,CAAC,sBAAsB,IAAI,wBAAwB,CAAC;AAEjE,MAAM,gBAAgB,GAAG,cAAc,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,OAAO,CAAC;AAC7D,MAAM,CAAC,MAAM,iBAAiB,GAC5B,OAAO,CAAC,GAAG,CAAC,iBAAiB,IAAI,gBAAgB,CAAC;AACpD,MAAM,CAAC,MAAM,qBAAqB,GAChC,OAAO,CAAC,GAAG,CAAC,qBAAqB,IAAI,4BAA4B,CAAC;AACpE,MAAM,CAAC,MAAM,yBAAyB,GACpC,OAAO,CAAC,GAAG,CAAC,yBAAyB,IAAI,aAAa,CAAC;AAEzD,MAAM,aAAa,GAAG,CAAC,CAAC;AAGxB,MAAM,mBAAmB,GAA2B;IAClD,4BAA4B,EAAE,WAAW;IACzC,uBAAuB,EAAE,WAAW;IACpC,SAAS,EAAE,WAAW;IACtB,uBAAuB,EAAE,cAAc;IACvC,kBAAkB,EAAE,cAAc;IAClC,YAAY,EAAE,cAAc;IAC5B,wBAAwB,EAAE,eAAe;IACzC,mBAAmB,EAAE,eAAe;IACpC,aAAa,EAAE,eAAe;IAC9B,uBAAuB,EAAE,eAAe;IACxC,kBAAkB,EAAE,eAAe;IACnC,aAAa,EAAE,eAAe;CAC/B,CAAC;AAEF,IAAI,aAAa,GAAkB,IAAI,CAAC;AAExC,SAAS,eAAe;IACtB,IAAI,CAAC,aAAa,EAAE,CAAC;QACnB,aAAa,GAAG,IAAI,MAAM,CAAC,EAAE,MAAM,EAAE,cAAc,EAAE,CAAC,CAAC;IACzD,CAAC;IACD,OAAO,aAAa,CAAC;AACvB,CAAC;AAED,SAAS,YAAY,CAAC,GAAY;IAChC,OAAO,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;AAC1D,CAAC;AAED,SAAS,kBAAkB,CAAC,SAAiB;IAC3C,OAAO,SAAS;SACb,IAAI,EAAE;SACN,OAAO,CAAC,oBAAoB,EAAE,EAAE,CAAC;SACjC,WAAW,EAAE;SACb,OAAO,CAAC,QAAQ,EAAE,GAAG,CAAC,CAAC;AAC5B,CAAC;AAED,SAAS,iBAAiB,CACxB,SAAiB,EACjB,cAAsC;IAEtC,MAAM,UAAU,GAAG,kBAAkB,CAAC,SAAS,CAAC,CAAC;IACjD,MAAM,QAAQ,GAAG,mBAAmB,CAAC,UAAU,CAAC,CAAC;IACjD,IAAI,QAAQ,IAAI,QAAQ,IAAI,cAAc;QAAE,OAAO,cAAc,CAAC,QAAQ,CAAC,CAAC;IAE5E,KAAK,MAAM,CAAC,GAAG,EAAE,KAAK,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,cAAc,CAAC,EAAE,CAAC;QAC1D,IAAI,GAAG,CAAC,WAAW,EAAE,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC,WAAW,EAAE;YAAE,OAAO,KAAK,CAAC;QACvE,IAAI,KAAK,CAAC,WAAW,EAAE,KAAK,UAAU;YAAE,OAAO,KAAK,CAAC;IACvD,CAAC;IAED,MAAM,SAAS,GAAG,MAAM,CAAC,MAAM,CAAC,cAAc,CAAC;SAC5C,MAAM,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,KAAK,cAAc,CAAC,MAAM,CAAC;SAClD,IAAI,CAAC,IAAI,CAAC,CAAC;IACd,MAAM,IAAI,KAAK,CACb,sCAAsC,SAAS,wBAAwB,SAAS,EAAE,CACnF,CAAC;AACJ,CAAC;AAkBD,MAAM,kBAAkB,GAAqC;IAC3D,MAAM,EAAE,EAAE,QAAQ,EAAE,IAAI,EAAE,WAAW,EAAE,GAAG,EAAE,WAAW,EAAE,GAAG,EAAE,SAAS,EAAE,IAAI,EAAE,aAAa,EAAE,IAAI,EAAE;IACpG,GAAG,EAAE,EAAE,QAAQ,EAAE,IAAI,EAAE,WAAW,EAAE,GAAG,EAAE,WAAW,EAAE,GAAG,EAAE,SAAS,EAAE,GAAG,EAAE,aAAa,EAAE,GAAG,EAAE;IAC/F,4EAA4E;IAC5E,2EAA2E;IAC3E,+EAA+E;IAC/E,8EAA8E;IAC9E,+EAA+E;IAC/E,6EAA6E;IAC7E,8EAA8E;IAC9E,EAAE,EAAE,EAAE,QAAQ,EAAE,IAAI,EAAE,WAAW,EAAE,KAAK,EAAE,WAAW,EAAE,IAAI,EAAE,SAAS,EAAE,IAAI,EAAE,aAAa,EAAE,GAAG,EAAE;IAClG,MAAM,EAAE,EAAE,QAAQ,EAAE,IAAI,EAAE,WAAW,EAAE,GAAG,EAAE,WAAW,EAAE,GAAG,EAAE,SAAS,EAAE,GAAG,EAAE,aAAa,EAAE,IAAI,EAAE;CACpG,CAAC;AAEF,IAAI,mBAAmB,GAAG,KAAK,CAAC;AAChC,SAAS,gBAAgB;IACvB,MAAM,KAAK,GAAG,mBAAmB,CAAC,kBAAkB,CAAC,qBAAqB,CAAC,CAAC,CAAC;IAC7E,IAAI,KAAK,KAAK,WAAW;QAAE,OAAO,IAAI,CAAC;IACvC,IAAI,KAAK,KAAK,eAAe;QAAE,OAAO,QAAQ,CAAC;IAC/C,IAAI,KAAK,KAAK,cAAc,IAAI,KAAK,KAAK,eAAe;QAAE,OAAO,KAAK,CAAC;IACxE,2EAA2E;IAC3E,+EAA+E;IAC/E,sEAAsE;IACtE,IAAI,CAAC,mBAAmB,EAAE,CAAC;QACzB,mBAAmB,GAAG,IAAI,CAAC;QAC3B,OAAO,CAAC,KAAK,CACX,8DAA8D;YAC5D,0BAA0B,qBAAqB,yBAAyB;YACxE,sEAAsE;YACtE,yFAAyF,CAC5F,CAAC;IACJ,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED,SAAS,YAAY,CAAC,IAAY;IAChC,MAAM,GAAG,GAAG,OAAO,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;IAC9B,IAAI,GAAG,KAAK,SAAS,IAAI,GAAG,CAAC,IAAI,EAAE,KAAK,EAAE;QAAE,OAAO,SAAS,CAAC;IAC7D,MAAM,CAAC,GAAG,MAAM,CAAC,GAAG,CAAC,CAAC;IACtB,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC;QAC1C,OAAO,CAAC,KAAK,CAAC,oCAAoC,IAAI,KAAK,GAAG,gCAAgC,CAAC,CAAC;QAChG,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,OAAO,CAAC,CAAC;AACX,CAAC;AAED,4EAA4E;AAC5E,+EAA+E;AAC/E,MAAM,UAAU,mBAAmB;IACjC,MAAM,IAAI,GACR,iBAAiB,KAAK,OAAO;QAC3B,CAAC,CAAC,kBAAkB,CAAC,MAAM;QAC3B,CAAC,CAAC,kBAAkB,CAAC,gBAAgB,EAAE,CAAC,CAAC;IAC7C,OAAO;QACL,QAAQ,EAAE,YAAY,CAAC,wBAAwB,CAAC,IAAI,IAAI,CAAC,QAAQ;QACjE,WAAW,EAAE,YAAY,CAAC,2BAA2B,CAAC,IAAI,IAAI,CAAC,WAAW;QAC1E,WAAW,EAAE,YAAY,CAAC,2BAA2B,CAAC,IAAI,IAAI,CAAC,WAAW;QAC1E,SAAS,EAAE,YAAY,CAAC,yBAAyB,CAAC,IAAI,IAAI,CAAC,SAAS;QACpE,aAAa,EAAE,YAAY,CAAC,6BAA6B,CAAC,IAAI,IAAI,CAAC,aAAa;KACjF,CAAC;AACJ,CAAC;AAED,8DAA8D;AAC9D,IAAI,WAAW,GAAQ,IAAI,CAAC;AAE5B,KAAK,UAAU,aAAa;IAC1B,IAAI,CAAC,WAAW,EAAE,CAAC;QACjB,IAAI,CAAC;YACH,MAAM,EAAE,aAAa,EAAE,cAAc,EAAE,GAAG,MAAM,MAAM,CAAC,WAAW,CAAC,CAAC;YACpE,MAAM,KAAK,GAAG,iBAAiB,CAAC,qBAAqB,EAAE,cAAc,CAAC,CAAC;YACvE,WAAW,GAAG,MAAM,aAAa,CAAC,IAAI,CAAC;gBACrC,KAAK,EAAE,KAAc;gBACrB,QAAQ,EAAE,yBAAyB;aACpC,CAAC,CAAC;QACL,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,MAAM,IAAI,KAAK,CACb,+CAA+C;gBAC7C,uCAAuC;gBACvC,mDAAmD;gBACnD,UAAU,YAAY,CAAC,GAAG,CAAC,EAAE,CAChC,CAAC;QACJ,CAAC;IACH,CAAC;IACD,OAAO,WAAW,CAAC;AACrB,CAAC;AAED,KAAK,UAAU,UAAU,CACvB,IAAY,EACZ,SAA6B;IAE7B,MAAM,KAAK,GAAG,MAAM,aAAa,EAAE,CAAC;IACpC,IAAI,SAAS,KAAK,OAAO,IAAI,OAAO,KAAK,CAAC,UAAU,KAAK,UAAU,EAAE,CAAC;QACpE,OAAO,KAAK,CAAC,IAAI,CAAC,MAAM,KAAK,CAAC,UAAU,CAAC,IAAI,CAAC,CAAa,CAAC;IAC9D,CAAC;IAED,MAAM,GAAG,GACP,OAAO,KAAK,CAAC,YAAY,KAAK,UAAU;QACtC,CAAC,CAAC,KAAK,CAAC,YAAY,CAAC,CAAC,IAAI,CAAC,EAAE,GAAG,CAAC;QACjC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC;IAC1B,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,GAAG,EAAE,CAAC;QAC9B,OAAO,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAa,CAAC;IAC1C,CAAC;IACD,MAAM,IAAI,KAAK,CAAC,kCAAkC,CAAC,CAAC;AACtD,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,cAAc,CAClC,IAAY,EACZ,YAAgC,SAAS;IAEzC,IAAI,iBAAiB,KAAK,OAAO,EAAE,CAAC;QAClC,OAAO,UAAU,CAAC,IAAI,EAAE,SAAS,CAAC,CAAC;IACrC,CAAC;IAED,MAAM,MAAM,GAAG,eAAe,EAAE,CAAC;IACjC,KAAK,IAAI,OAAO,GAAG,CAAC,EAAE,OAAO,GAAG,aAAa,EAAE,OAAO,EAAE,EAAE,CAAC;QACzD,IAAI,CAAC;YACH,MAAM,IAAI,GAAG,MAAM,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC;gBAC1C,KAAK,EAAE,sBAAsB;gBAC7B,KAAK,EAAE,IAAI;aACZ,CAAC,CAAC;YACH,OAAO,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;QAChC,CAAC;QAAC,OAAO,GAAG,EAAE,CAAC;YACb,IAAI,OAAO,KAAK,aAAa,GAAG,CAAC;gBAAE,MAAM,GAAG,CAAC;YAC7C,MAAM,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,EAAE,CAC5B,UAAU,CAAC,OAAO,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,OAAO,CAAC,GAAG,IAAI,CAAC,CACjD,CAAC;QACJ,CAAC;IACH,CAAC;IACD,MAAM,IAAI,KAAK,CAAC,aAAa,CAAC,CAAC;AACjC,CAAC"}

package/dist/memoryGraph.d.ts

@@ -9,19 +9,29 @@ export declare class MemoryGraph {
     private _storedDim;
     private _lock;
     private _dirty;
-    static readonly HOP_DECAY = 0.5;
+    private _bm25;
+    constructor();
+    static readonly HOP_DECAY = 0.3;
     static readonly TIME_HALF_LIFE: number;
     get linkCount(): number;
     private _link;
     private _hasLink;
+    private _getLinkWeight;
+    private _setLinkWeight;
     private _unlinkMemory;
+    private _validMemoryLinks;
+    private _removeMemoryReferences;
+    private _pruneDanglingExplicitLinks;
     private _pruneOrphanKeys;
     private _checkDim;
+    private _ensureEmbeddingDim;
+    private _migrateEmbeddings;
     private _isExpired;
     private _timeFactor;
     private _keyIdf;
     private _findDuplicate;
     private _autoLinkKeys;
+    private _rebuildBm25Index;
     getKeysForMemory(memId: string): string[];
     load(): Promise<void>;
     save(): Promise<void>;

package/dist/memoryGraph.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"memoryGraph.d.ts","sourceRoot":"","sources":["../src/memoryGraph.ts"],"names":[],"mappings":"AAMA,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,EAAa,MAAM,YAAY,CAAC;AA0CzD,wBAAgB,YAAY,CAAC,IAAI,EAAE,OAAO,GAAG,MAAM,EAAE,CAgBpD;AAID,qBAAa,WAAW;IACtB,IAAI,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,CAAM;IAC/B,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAM;IAEtC,OAAO,CAAC,UAAU,CAAmC;IACrD,OAAO,CAAC,UAAU,CAAmC;IACrD,OAAO,CAAC,aAAa,CAA8B;IACnD,OAAO,CAAC,UAAU,CAAuB;IACzC,OAAO,CAAC,KAAK,CAAe;IAC5B,OAAO,CAAC,MAAM,CAAS;IAEvB,MAAM,CAAC,QAAQ,CAAC,SAAS,OAAO;IAChC,MAAM,CAAC,QAAQ,CAAC,cAAc,SAAkB;IAEhD,IAAI,SAAS,IAAI,MAAM,CAKtB;IAED,OAAO,CAAC,KAAK;IAOb,OAAO,CAAC,QAAQ;IAIhB,OAAO,CAAC,aAAa;IAcrB,OAAO,CAAC,gBAAgB;IAOxB,OAAO,CAAC,SAAS;IAejB,OAAO,CAAC,UAAU;IAIlB,OAAO,CAAC,WAAW;IAOnB,OAAO,CAAC,OAAO;IASf,OAAO,CAAC,cAAc;IAkBtB,OAAO,CAAC,aAAa;IAYrB,gBAAgB,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE;IAUnC,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;IAoDrB,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;IAiB3B,SAAS,IAAI,IAAI;IAIX,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAMtB,eAAe,CACnB,OAAO,EAAE,MAAM,EACf,OAAO,GAAE,SAAS,GAAG,MAAM,GAAG,aAAyB,GACtD,OAAO,CAAC,MAAM,CAAC;IAwCZ,GAAG,CACP,OAAO,EAAE,MAAM,EACf,WAAW,EAAE,MAAM,EAAE,EACrB,OAAO,GAAE;QACP,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,IAAI,CAAC;QACzC,MAAM,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,IAAI,CAAC;QACxC,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,UAAU,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;QAC3B,SAAS,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;KACxB,GACL,OAAO,CAAC,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAkEvB,SAAS,CACb,KAAK,EAAE,MAAM,EACb,UAAU,EAAE,MAAM,EAClB,OAAO,GAAE;QACP,WAAW,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;QAC9B,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,IAAI,CAAC;QACzC,MAAM,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,IAAI,CAAC;QACxC,SAAS,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;QAC1B,SAAS,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;KACxB,GACL,OAAO,CAAC,MAAM,CAAC;IA8EZ,MAAM,CACV,KAAK,EAAE,MAAM,EACb,IAAI,SAAI,EACR,SAAS,CAAC,EAAE,MAAM,GAAG,IAAI,EACzB,MAAM,UAAQ,GACb,OAAO,CAAC,MAAM,EAAE,CAAC;IAsKpB,UAAU,CAAC,QAAQ,EAAE,MAAM,GAAG,MAAM,EAAE;IAiFhC,MAAM,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC,OAAO,CAAC;IAiBhD,OAAO,CAAC,SAAS,CAAC,EAAE,MAAM,GAAG,IAAI,GAAG,MAAM,EAAE;IAwBtC,cAAc,IAAI,OAAO,CAAC,MAAM,CAAC;CAmBxC;AAID,wBAAsB,QAAQ,CAC5B,SAAS,EAAE,MAAM,EACjB,IAAI,EAAE,MAAM,EACZ,OAAO,EAAE,MAAM,GACd,OAAO,CAAC,MAAM,CAAC,CAkBjB;AAED,wBAAsB,gBAAgB,CACpC,SAAS,EAAE,MAAM,EACjB,IAAI,CAAC,EAAE,MAAM,GAAG,IAAI,GACnB,OAAO,CAAC,MAAM,EAAE,CAAC,CAkBnB"}
+{"version":3,"file":"memoryGraph.d.ts","sourceRoot":"","sources":["../src/memoryGraph.ts"],"names":[],"mappings":"AAOA,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,EAAa,MAAM,YAAY,CAAC;AAwEzD,wBAAgB,YAAY,CAAC,IAAI,EAAE,OAAO,GAAG,MAAM,EAAE,CAgBpD;AAID,qBAAa,WAAW;IACtB,IAAI,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,CAAM;IAC/B,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAM;IAEtC,OAAO,CAAC,UAAU,CAA2C;IAC7D,OAAO,CAAC,UAAU,CAA2C;IAC7D,OAAO,CAAC,aAAa,CAA8B;IACnD,OAAO,CAAC,UAAU,CAAuB;IACzC,OAAO,CAAC,KAAK,CAAe;IAC5B,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,KAAK,CAAa;;IAgB1B,MAAM,CAAC,QAAQ,CAAC,SAAS,OAAO;IAChC,MAAM,CAAC,QAAQ,CAAC,cAAc,SAAkB;IAEhD,IAAI,SAAS,IAAI,MAAM,CAKtB;IAED,OAAO,CAAC,KAAK;IAWb,OAAO,CAAC,QAAQ;IAIhB,OAAO,CAAC,cAAc;IAItB,OAAO,CAAC,cAAc;IAMtB,OAAO,CAAC,aAAa;IAcrB,OAAO,CAAC,iBAAiB;IAczB,OAAO,CAAC,uBAAuB;IAS/B,OAAO,CAAC,2BAA2B;IAMnC,OAAO,CAAC,gBAAgB;IAOxB,OAAO,CAAC,SAAS;YAqBH,mBAAmB;YAsBnB,kBAAkB;IAwBhC,OAAO,CAAC,UAAU;IAIlB,OAAO,CAAC,WAAW;IAOnB,OAAO,CAAC,OAAO;IASf,OAAO,CAAC,cAAc;IAkBtB,OAAO,CAAC,aAAa;IAYrB,OAAO,CAAC,iBAAiB;IASzB,gBAAgB,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE;IAUnC,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;IAgErB,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC;IAmB3B,SAAS,IAAI,IAAI;IAIX,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAMtB,eAAe,CACnB,OAAO,EAAE,MAAM,EACf,OAAO,GAAE,SAAS,GAAG,MAAM,GAAG,aAAyB,GACtD,OAAO,CAAC,MAAM,CAAC;IAwCZ,GAAG,CACP,OAAO,EAAE,MAAM,EACf,WAAW,EAAE,MAAM,EAAE,EACrB,OAAO,GAAE;QACP,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,IAAI,CAAC;QACzC,MAAM,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,IAAI,CAAC;QACxC,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,UAAU,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;QAC3B,SAAS,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;KACxB,GACL,OAAO,CAAC,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAiEvB,SAAS,CACb,KAAK,EAAE,MAAM,EACb,UAAU,EAAE,MAAM,EAClB,OAAO,GAAE;QACP,WAAW,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;QAC9B,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,GAAG,IAAI,CAAC;QACzC,MAAM,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,IAAI,CAAC;QACxC,SAAS,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;QAC1B,SAAS,CAAC,EAAE,MAAM,EAAE,GAAG,IAAI,CAAC;KACxB,GACL,OAAO,CAAC,MAAM,CAAC;IAkFZ,MAAM,CACV,KAAK,EAAE,MAAM,EACb,IAAI,SAAI,EACR,SAAS,CAAC,EAAE,MAAM,GAAG,IAAI,EACzB,MAAM,UAAQ,GACb,OAAO,CAAC,MAAM,EAAE,CAAC;IA0PpB,UAAU,CAAC,QAAQ,EAAE,MAAM,GAAG,MAAM,EAAE;IAiFhC,MAAM,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC,OAAO,CAAC;IAmBhD,OAAO,CAAC,SAAS,CAAC,EAAE,MAAM,GAAG,IAAI,GAAG,MAAM,EAAE;IAwBtC,cAAc,IAAI,OAAO,CAAC,MAAM,CAAC;CAqBxC;AAID,wBAAsB,QAAQ,CAC5B,SAAS,EAAE,MAAM,EACjB,IAAI,EAAE,MAAM,EACZ,OAAO,EAAE,MAAM,GACd,OAAO,CAAC,MAAM,CAAC,CAkBjB;AAED,wBAAsB,gBAAgB,CACpC,SAAS,EAAE,MAAM,EACjB,IAAI,CAAC,EAAE,MAAM,GAAG,IAAI,GACnB,OAAO,CAAC,MAAM,EAAE,CAAC,CAyBnB"}