@yamo/memory-mesh 3.0.0 → 3.0.1

package/lib/memory/schema.ts

@@ -1,3 +1,4 @@
+// @ts-nocheck
 /**
  * LanceDB Schema Definitions for MemoryManager
  * Uses Apache Arrow Schema format for LanceDB JavaScript SDK
@@ -7,197 +8,130 @@
  * - all-mpnet-base-v2: 768 dimensions
  * - text-embedding-3-small: 1536 dimensions
  */
-
 import * as arrow from "apache-arrow";
-import * as lancedb from "@lancedb/lancedb";
-
 /**
  * Default vector dimension (all-MiniLM-L6-v2)
  */
 export const DEFAULT_VECTOR_DIMENSION = 384;
-
 /**
  * Common embedding model dimensions
  */
-export const EMBEDDING_DIMENSIONS: Record<string, number> = {
-  "Xenova/all-MiniLM-L6-v2": 384,
-  "Xenova/all-mpnet-base-v2": 768,
-  "Xenova/distiluse-base-multilingual-cased-v1": 512,
-  "sentence-transformers/all-MiniLM-L6-v2": 384,
-  "sentence-transformers/all-mpnet-base-v2": 768,
-  "openai/text-embedding-3-small": 1536,
-  "openai/text-embedding-3-large": 3072,
-  "cohere/embed-english-light-v3.0": 1024,
-  "cohere/embed-english-v3.0": 1024,
+export const EMBEDDING_DIMENSIONS = {
+    "Xenova/all-MiniLM-L6-v2": 384,
+    "Xenova/all-mpnet-base-v2": 768,
+    "Xenova/distiluse-base-multilingual-cased-v1": 512,
+    "sentence-transformers/all-MiniLM-L6-v2": 384,
+    "sentence-transformers/all-mpnet-base-v2": 768,
+    "openai/text-embedding-3-small": 1536,
+    "openai/text-embedding-3-large": 3072,
+    "cohere/embed-english-light-v3.0": 1024,
+    "cohere/embed-english-v3.0": 1024,
 };
-
 /**
  * Get dimension for a given embedding model
  * @param {string} modelName - Embedding model name or path
  * @returns {number} Vector dimension
  */
-export function getEmbeddingDimension(modelName?: string): number {
-  if (!modelName) {
-    return DEFAULT_VECTOR_DIMENSION;
-  }
-
-  // Check exact match
-  if (EMBEDDING_DIMENSIONS[modelName]) {
-    return EMBEDDING_DIMENSIONS[modelName];
-  }
-
-  // Check for partial matches
-  for (const [key, dimension] of Object.entries(EMBEDDING_DIMENSIONS)) {
-    if (modelName.toLowerCase().includes(key.toLowerCase())) {
-      return dimension;
+export function getEmbeddingDimension(modelName) {
+    if (!modelName) {
+        return DEFAULT_VECTOR_DIMENSION;
+    }
+    // Check exact match
+    if (EMBEDDING_DIMENSIONS[modelName]) {
+        return EMBEDDING_DIMENSIONS[modelName];
     }
-  }
-
-  // Fallback to default
-  return DEFAULT_VECTOR_DIMENSION;
+    // Check for partial matches
+    for (const [key, dimension] of Object.entries(EMBEDDING_DIMENSIONS)) {
+        if (modelName.toLowerCase().includes(key.toLowerCase())) {
+            return dimension;
+        }
+    }
+    // Fallback to default
+    return DEFAULT_VECTOR_DIMENSION;
 }
-
 /**
  * Create a memory schema with a specific vector dimension
  * @param {number} vectorDim - Vector dimension (e.g., 384, 768, 1536)
  * @returns {arrow.Schema} Arrow schema with specified dimension
  */
-export function createMemorySchema(
-  vectorDim: number = DEFAULT_VECTOR_DIMENSION,
-): arrow.Schema {
-  return new arrow.Schema([
-    new arrow.Field("id", new arrow.Utf8(), false),
-    new arrow.Field(
-      "vector",
-      new arrow.FixedSizeList(
-        vectorDim,
-        new arrow.Field("item", new arrow.Float32(), true),
-      ),
-      false,
-    ),
-    new arrow.Field("content", new arrow.Utf8(), false),
-    new arrow.Field("metadata", new arrow.Utf8(), true), // Stored as JSON string
-    new arrow.Field(
-      "created_at",
-      new arrow.Timestamp(arrow.TimeUnit.MILLISECOND),
-      false,
-    ),
-    new arrow.Field(
-      "updated_at",
-      new arrow.Timestamp(arrow.TimeUnit.MILLISECOND),
-      true,
-    ),
-  ]);
+export function createMemorySchema(vectorDim = DEFAULT_VECTOR_DIMENSION) {
+    return new arrow.Schema([
+        new arrow.Field("id", new arrow.Utf8(), false),
+        new arrow.Field("vector", new arrow.FixedSizeList(vectorDim, new arrow.Field("item", new arrow.Float32(), true)), false),
+        new arrow.Field("content", new arrow.Utf8(), false),
+        new arrow.Field("metadata", new arrow.Utf8(), true), // Stored as JSON string
+        new arrow.Field("created_at", new arrow.Timestamp(arrow.TimeUnit.MILLISECOND), false),
+        new arrow.Field("updated_at", new arrow.Timestamp(arrow.TimeUnit.MILLISECOND), true),
+    ]);
 }
-
 /**
  * Create V2 memory schema with automatic recall fields
  * All new fields are nullable for backward compatibility
  * @param {number} vectorDim - Vector dimension (e.g., 384, 768, 1536)
  * @returns {arrow.Schema} Arrow schema with V2 fields
  */
-export function createMemorySchemaV2(
-  vectorDim: number = DEFAULT_VECTOR_DIMENSION,
-): arrow.Schema {
-  return new arrow.Schema([
-    // ========== V1 Fields (Backward Compatible) ==========
-    new arrow.Field("id", new arrow.Utf8(), false),
-    new arrow.Field(
-      "vector",
-      new arrow.FixedSizeList(
-        vectorDim,
-        new arrow.Field("item", new arrow.Float32(), true),
-      ),
-      false,
-    ),
-    new arrow.Field("content", new arrow.Utf8(), false),
-    new arrow.Field("metadata", new arrow.Utf8(), true),
-    new arrow.Field(
-      "created_at",
-      new arrow.Timestamp(arrow.TimeUnit.MILLISECOND),
-      false,
-    ),
-    new arrow.Field(
-      "updated_at",
-      new arrow.Timestamp(arrow.TimeUnit.MILLISECOND),
-      true,
-    ),
-
-    // ========== V2 Fields (All Nullable) ==========
-    new arrow.Field("session_id", new arrow.Utf8(), true), // Session association
-    new arrow.Field("agent_id", new arrow.Utf8(), true), // Agent/skill that created memory
-    new arrow.Field("memory_type", new arrow.Utf8(), true), // 'global', 'session', 'agent'
-    new arrow.Field("importance_score", new arrow.Float32(), true), // 0.0-1.0 importance
-    new arrow.Field("access_count", new arrow.Int32(), true), // Popularity tracking
-    new arrow.Field(
-      "last_accessed",
-      new arrow.Timestamp(arrow.TimeUnit.MILLISECOND),
-      true,
-    ),
-  ]);
+export function createMemorySchemaV2(vectorDim = DEFAULT_VECTOR_DIMENSION) {
+    return new arrow.Schema([
+        // ========== V1 Fields (Backward Compatible) ==========
+        new arrow.Field("id", new arrow.Utf8(), false),
+        new arrow.Field("vector", new arrow.FixedSizeList(vectorDim, new arrow.Field("item", new arrow.Float32(), true)), false),
+        new arrow.Field("content", new arrow.Utf8(), false),
+        new arrow.Field("metadata", new arrow.Utf8(), true),
+        new arrow.Field("created_at", new arrow.Timestamp(arrow.TimeUnit.MILLISECOND), false),
+        new arrow.Field("updated_at", new arrow.Timestamp(arrow.TimeUnit.MILLISECOND), true),
+        // ========== V2 Fields (All Nullable) ==========
+        new arrow.Field("session_id", new arrow.Utf8(), true), // Session association
+        new arrow.Field("agent_id", new arrow.Utf8(), true), // Agent/skill that created memory
+        new arrow.Field("memory_type", new arrow.Utf8(), true), // 'global', 'session', 'agent'
+        new arrow.Field("importance_score", new arrow.Float32(), true), // 0.0-1.0 importance
+        new arrow.Field("access_count", new arrow.Int32(), true), // Popularity tracking
+        new arrow.Field("last_accessed", new arrow.Timestamp(arrow.TimeUnit.MILLISECOND), true),
+    ]);
 }
-
 /**
  * Create schema for synthesized skills (Recursive Skill Synthesis)
  * @param {number} vectorDim - Vector dimension for intent embedding
  * @returns {arrow.Schema} Arrow schema
  */
-export function createSynthesizedSkillSchema(
-  vectorDim: number = DEFAULT_VECTOR_DIMENSION,
-): arrow.Schema {
-  return new arrow.Schema([
-    new arrow.Field("id", new arrow.Utf8(), false),
-    new arrow.Field("name", new arrow.Utf8(), false),
-    new arrow.Field("intent", new arrow.Utf8(), false),
-    new arrow.Field("yamo_text", new arrow.Utf8(), false),
-    new arrow.Field(
-      "vector",
-      new arrow.FixedSizeList(
-        vectorDim,
-        new arrow.Field("item", new arrow.Float32(), true),
-      ),
-      false,
-    ),
-    new arrow.Field("metadata", new arrow.Utf8(), true), // Stored as JSON: {reliability, use_count, created_at}
-    new arrow.Field(
-      "created_at",
-      new arrow.Timestamp(arrow.TimeUnit.MILLISECOND),
-      false,
-    ),
-  ]);
+export function createSynthesizedSkillSchema(vectorDim = DEFAULT_VECTOR_DIMENSION) {
+    return new arrow.Schema([
+        new arrow.Field("id", new arrow.Utf8(), false),
+        new arrow.Field("name", new arrow.Utf8(), false),
+        new arrow.Field("intent", new arrow.Utf8(), false),
+        new arrow.Field("yamo_text", new arrow.Utf8(), false),
+        new arrow.Field("vector", new arrow.FixedSizeList(vectorDim, new arrow.Field("item", new arrow.Float32(), true)), false),
+        new arrow.Field("metadata", new arrow.Utf8(), true), // Stored as JSON: {reliability, use_count, created_at}
+        new arrow.Field("created_at", new arrow.Timestamp(arrow.TimeUnit.MILLISECOND), false),
+    ]);
 }
-
 /**
  * Check if a table is using V2 schema
  * @param {arrow.Schema} schema - Table schema to check
  * @returns {boolean} True if V2 schema detected
  */
-export function isSchemaV2(schema: arrow.Schema): boolean {
-  return schema.fields.some((f) => f.name === "session_id");
+export function isSchemaV2(schema) {
+    return schema.fields.some((f) => f.name === "session_id");
 }
-
 /**
  * Memory table schema using Apache Arrow format (default 384 dimensions)
  * @deprecated Use createMemorySchema(vectorDim) for dynamic dimensions
  */
 export const MEMORY_SCHEMA = createMemorySchema(DEFAULT_VECTOR_DIMENSION);
-
 /**
  * Index configuration for memory table
  * Indices should be created after data is inserted
  */
 export const INDEX_CONFIG = {
-  vector: {
-    index_type: "ivf_pq",
-    metric: "cosine",
-    num_partitions: 256,
-    num_sub_vectors: 8,
-  },
-  full_text: {
-    fields: ["content"],
-  },
+    vector: {
+        index_type: "ivf_pq",
+        metric: "cosine",
+        num_partitions: 256,
+        num_sub_vectors: 8,
+    },
+    full_text: {
+        fields: ["content"],
+    },
 };
-
 /**
  * Creates a memory table in LanceDB with the predefined schema (384 dimensions)
  * @param {lancedb.Connection} db - LanceDB connection
@@ -206,17 +140,9 @@ export const INDEX_CONFIG = {
  * @throws {Error} If table creation fails
  * @deprecated Use createMemoryTableWithDimension() for dynamic dimensions
  */
-export async function createMemoryTable(
-  db: lancedb.Connection,
-  tableName: string = "memory_entries",
-): Promise<lancedb.Table> {
-  return createMemoryTableWithDimension(
-    db,
-    tableName,
-    DEFAULT_VECTOR_DIMENSION,
-  );
+export async function createMemoryTable(db, tableName = "memory_entries") {
+    return createMemoryTableWithDimension(db, tableName, DEFAULT_VECTOR_DIMENSION);
 }
-
 /**
  * Creates a memory table in LanceDB with a specific vector dimension
  * @param {lancedb.Connection} db - LanceDB connection
@@ -225,43 +151,34 @@ export async function createMemoryTable(
  * @returns {Promise<lancedb.Table>} The created or opened table
  * @throws {Error} If table creation fails
  */
-export async function createMemoryTableWithDimension(
-  db: lancedb.Connection,
-  tableName: string,
-  vectorDim: number,
-): Promise<lancedb.Table> {
-  try {
-    // Check if table already exists
-    const existingTables = await db.tableNames();
-
-    if (existingTables.includes(tableName)) {
-      return await db.openTable(tableName);
+export async function createMemoryTableWithDimension(db, tableName, vectorDim) {
+    try {
+        // Check if table already exists
+        const existingTables = await db.tableNames();
+        if (existingTables.includes(tableName)) {
+            return await db.openTable(tableName);
+        }
+        // Create schema with specified dimension
+        const schema = createMemorySchema(vectorDim);
+        // Create table with schema
+        // LanceDB v0.23.0+ accepts empty array as initial data with schema option
+        const table = await db.createTable(tableName, [], { schema }); // Cast to any because lancedb types might be strict about options
+        return table;
+    }
+    catch (error) {
+        const message = error instanceof Error ? error.message : String(error);
+        throw new Error(`Failed to create memory table with dimension ${vectorDim}: ${message}`);
     }
-
-    // Create schema with specified dimension
-    const schema = createMemorySchema(vectorDim);
-
-    // Create table with schema
-    // LanceDB v0.23.0+ accepts empty array as initial data with schema option
-    const table = await db.createTable(tableName, [], { schema } as any); // Cast to any because lancedb types might be strict about options
-    return table;
-  } catch (error) {
-    const message = error instanceof Error ? error.message : String(error);
-    throw new Error(
-      `Failed to create memory table with dimension ${vectorDim}: ${message}`,
-    );
-  }
 }
-
 export default {
-  MEMORY_SCHEMA,
-  INDEX_CONFIG,
-  createMemoryTable,
-  createMemoryTableWithDimension,
-  createMemorySchema,
-  createMemorySchemaV2,
-  isSchemaV2,
-  getEmbeddingDimension,
-  DEFAULT_VECTOR_DIMENSION,
-  EMBEDDING_DIMENSIONS,
+    MEMORY_SCHEMA,
+    INDEX_CONFIG,
+    createMemoryTable,
+    createMemoryTableWithDimension,
+    createMemorySchema,
+    createMemorySchemaV2,
+    isSchemaV2,
+    getEmbeddingDimension,
+    DEFAULT_VECTOR_DIMENSION,
+    EMBEDDING_DIMENSIONS,
 };

package/lib/memory/scorer.d.ts

@@ -1,26 +1,25 @@
 /**
  * MemoryScorer - Calculate memory importance and detect duplicates
  */
-import { MemoryMesh } from "./memory-mesh.js";
 export declare class MemoryScorer {
     #private;
     /**
      * @param {MemoryMesh} mesh - MemoryMesh instance for duplicate checking
      */
-    constructor(mesh: MemoryMesh);
+    constructor(mesh: any);
     /**
      * Calculate importance score for content
      * @param {string} content - Content to score
      * @param {Object} metadata - Associated metadata
      * @returns {Promise<number>} Importance score (0-1)
      */
-    calculateImportance(content: string, metadata?: any): number;
+    calculateImportance(content: any, metadata?: {}): number;
     /**
      * Check if content is duplicate of existing memory
      * @param {string} content - Content to check
      * @param {number} threshold - Similarity threshold (default 0.9)
      * @returns {Promise<boolean>} True if duplicate exists
      */
-    isDuplicate(content: string, threshold?: number): Promise<boolean>;
+    isDuplicate(content: any, threshold?: number): Promise<boolean>;
 }
 export default MemoryScorer;

package/lib/memory/scorer.js

@@ -1,3 +1,4 @@
+// @ts-nocheck
 /**
  * MemoryScorer - Calculate memory importance and detect duplicates
  */

package/lib/memory/scorer.ts

@@ -1,95 +1,78 @@
+// @ts-nocheck
 /**
  * MemoryScorer - Calculate memory importance and detect duplicates
  */
-
-import { MemoryMesh } from "./memory-mesh.js";
-
 export class MemoryScorer {
-  #mesh: MemoryMesh;
-
-  /**
-   * @param {MemoryMesh} mesh - MemoryMesh instance for duplicate checking
-   */
-  constructor(mesh: MemoryMesh) {
-    this.#mesh = mesh;
-  }
-
-  /**
-   * Calculate importance score for content
-   * @param {string} content - Content to score
-   * @param {Object} metadata - Associated metadata
-   * @returns {Promise<number>} Importance score (0-1)
-   */
-  calculateImportance(content: string, metadata: any = {}): number {
-    let score = 0;
-
-    // Content length (longer = more important, up to a point)
-    const length = content.length;
-    score += Math.min(length / 1000, 0.2);
-
-    // Has structured data (JSON, code blocks)
-    if (content.includes("```") || content.includes("{")) {
-      score += 0.1;
+    #mesh;
+    /**
+     * @param {MemoryMesh} mesh - MemoryMesh instance for duplicate checking
+     */
+    constructor(mesh) {
+        this.#mesh = mesh;
     }
-
-    // Interaction type bonuses
-    if (metadata.interaction_type === "tool_execution") {
-      score += 0.15;
+    /**
+     * Calculate importance score for content
+     * @param {string} content - Content to score
+     * @param {Object} metadata - Associated metadata
+     * @returns {Promise<number>} Importance score (0-1)
+     */
+    calculateImportance(content, metadata = {}) {
+        let score = 0;
+        // Content length (longer = more important, up to a point)
+        const length = content.length;
+        score += Math.min(length / 1000, 0.2);
+        // Has structured data (JSON, code blocks)
+        if (content.includes("```") || content.includes("{")) {
+            score += 0.1;
+        }
+        // Interaction type bonuses
+        if (metadata.interaction_type === "tool_execution") {
+            score += 0.15;
+        }
+        if (metadata.interaction_type === "file_operation") {
+            score += 0.1;
+        }
+        // Tool usage indicates importance
+        if (metadata.tools_used?.length > 0) {
+            score += Math.min(metadata.tools_used.length * 0.05, 0.15);
+        }
+        // File involvement
+        if (metadata.files_involved?.length > 0) {
+            score += Math.min(metadata.files_involved.length * 0.05, 0.15);
+        }
+        // Keywords that indicate importance
+        const importantKeywords = [
+            "error",
+            "bug",
+            "fix",
+            "important",
+            "critical",
+            "note",
+            "remember",
+        ];
+        const lowerContent = content.toLowerCase();
+        const keywordMatches = importantKeywords.filter((k) => lowerContent.includes(k)).length;
+        score += Math.min(keywordMatches * 0.05, 0.15);
+        return Math.min(score, 1.0);
     }
-    if (metadata.interaction_type === "file_operation") {
-      score += 0.1;
+    /**
+     * Check if content is duplicate of existing memory
+     * @param {string} content - Content to check
+     * @param {number} threshold - Similarity threshold (default 0.9)
+     * @returns {Promise<boolean>} True if duplicate exists
+     */
+    async isDuplicate(content, threshold = 0.9) {
+        try {
+            const results = await this.#mesh.search(content, {
+                limit: 1,
+                useCache: false,
+            });
+            return results.length > 0 && results[0].score >= threshold;
+        }
+        catch (_error) {
+            // On error, assume not duplicate to allow storage
+            return false;
+        }
     }
-
-    // Tool usage indicates importance
-    if (metadata.tools_used?.length > 0) {
-      score += Math.min(metadata.tools_used.length * 0.05, 0.15);
-    }
-
-    // File involvement
-    if (metadata.files_involved?.length > 0) {
-      score += Math.min(metadata.files_involved.length * 0.05, 0.15);
-    }
-
-    // Keywords that indicate importance
-    const importantKeywords = [
-      "error",
-      "bug",
-      "fix",
-      "important",
-      "critical",
-      "note",
-      "remember",
-    ];
-    const lowerContent = content.toLowerCase();
-    const keywordMatches = importantKeywords.filter((k) =>
-      lowerContent.includes(k),
-    ).length;
-    score += Math.min(keywordMatches * 0.05, 0.15);
-
-    return Math.min(score, 1.0);
-  }
-
-  /**
-   * Check if content is duplicate of existing memory
-   * @param {string} content - Content to check
-   * @param {number} threshold - Similarity threshold (default 0.9)
-   * @returns {Promise<boolean>} True if duplicate exists
-   */
-  async isDuplicate(
-    content: string,
-    threshold: number = 0.9,
-  ): Promise<boolean> {
-    try {
-      const results = await this.#mesh.search(content, {
-        limit: 1,
-        useCache: false,
-      });
-      return results.length > 0 && results[0].score >= threshold;
-    } catch (_error) {
-      // On error, assume not duplicate to allow storage
-      return false;
-    }
-  }
 }
-
 export default MemoryScorer;

package/lib/memory/search/index.js

@@ -1 +1,2 @@
+// @ts-nocheck
 export { KeywordSearch } from "./keyword-search.js";

package/lib/memory/search/index.ts

@@ -1 +1,2 @@
+// @ts-nocheck
 export { KeywordSearch } from "./keyword-search.js";

package/lib/memory/search/keyword-search.d.ts

@@ -2,43 +2,31 @@
  * Simple Keyword Search Engine (In-Memory)
  * Provides basic TF-IDF style retrieval to complement vector search
  */
-export interface KeywordDoc {
-    content: string;
-    metadata?: any;
-}
-export interface KeywordSearchResult extends KeywordDoc {
-    id: string;
-    score: number;
-    matches: string[];
-}
-export interface SearchOptions {
-    limit?: number;
-}
 export declare class KeywordSearch {
-    index: Map<string, Map<string, number>>;
-    docLengths: Map<string, number>;
-    idf: Map<string, number>;
-    docs: Map<string, KeywordDoc>;
-    isDirty: boolean;
+    index: any;
+    docLengths: any;
+    idf: any;
+    docs: any;
+    isDirty: any;
     constructor();
     /**
      * Tokenize text into normalized terms
      * @param {string} text
      * @returns {string[]} tokens
      */
-    tokenize(text: string): string[];
+    tokenize(text: any): any;
     /**
      * Add a document to the index
      * @param {string} id
      * @param {string} content
      * @param {Object} [metadata]
      */
-    add(id: string, content: string, metadata?: any): void;
+    add(id: any, content: any, metadata?: {}): void;
     /**
      * Remove a document
      * @param {string} id
      */
-    remove(id: string): void;
+    remove(id: any): void;
     /**
      * Recalculate IDF scores
      */
@@ -49,14 +37,10 @@ export declare class KeywordSearch {
      * @param {Object} options
      * @returns {Array<{id: string, score: number, matches: string[], content: string, metadata: Object}>}
      */
-    search(query: string, options?: SearchOptions): KeywordSearchResult[];
+    search(query: any, options?: {}): any[];
     /**
      * Bulk load records
      * @param {Array} records
      */
-    load(records: {
-        id: string;
-        content: string;
-        metadata?: any;
-    }[]): void;
+    load(records: any): void;
 }

package/lib/memory/search/keyword-search.js

@@ -1,3 +1,4 @@
+// @ts-nocheck
 /**
  * Simple Keyword Search Engine (In-Memory)
  * Provides basic TF-IDF style retrieval to complement vector search