@memvid/sdk 2.0.152 → 2.0.153

package/dist/embeddings.d.ts

@@ -311,10 +311,70 @@ export declare class MistralEmbeddings implements EmbeddingProvider {
     embedDocuments(texts: string[]): Promise<number[][]>;
     embedQuery(text: string): Promise<number[]>;
 }
+/**
+ * Ollama embedding provider configuration.
+ */
+export interface OllamaEmbeddingsConfig {
+    /** Ollama server URL. Default: 'http://localhost:11434' or OLLAMA_HOST env var */
+    baseUrl?: string;
+    /** Model to use. Default: 'nomic-embed-text' */
+    model?: string;
+    /** Known embedding dimension. If omitted, auto-detected on first call. */
+    dimension?: number;
+}
+/**
+ * Ollama embedding provider.
+ *
+ * Uses a local Ollama server to generate embeddings. Supports any embedding model
+ * available in Ollama, including nomic-embed-text, mxbai-embed-large, all-minilm, etc.
+ *
+ * @example
+ * ```typescript
+ * // Default: localhost:11434 with nomic-embed-text
+ * const embedder = new OllamaEmbeddings();
+ *
+ * // Custom configuration
+ * const embedder = new OllamaEmbeddings({
+ *   baseUrl: 'http://gpu-server:11434',
+ *   model: 'nomic-embed-text',
+ * });
+ *
+ * // Use with Memvid
+ * const texts = ['Hello world', 'Goodbye world'];
+ * const embeddings = await embedder.embedDocuments(texts);
+ *
+ * // Or embed and store in one step
+ * const embedding = await embedder.embedQuery('Document text...');
+ * await mem.put({
+ *   title: 'My Doc',
+ *   label: 'docs',
+ *   text: 'Document text...',
+ *   embedding,
+ *   embeddingIdentity: {
+ *     provider: 'ollama',
+ *     model: 'nomic-embed-text',
+ *     dimension: embedding.length,
+ *   },
+ * });
+ * ```
+ */
+export declare class OllamaEmbeddings implements EmbeddingProvider {
+    private readonly _baseUrl;
+    private readonly _model;
+    private _dimension?;
+    constructor(config?: OllamaEmbeddingsConfig);
+    private static readonly OLLAMA_MODEL_DIMENSIONS;
+    get dimension(): number;
+    get modelName(): string;
+    get provider(): string;
+    private setDimensionFromEmbedding;
+    embedDocuments(texts: string[]): Promise<number[][]>;
+    embedQuery(text: string): Promise<number[]>;
+}
 /**
  * Factory function to create an embedding provider.
  *
- * @param provider - One of: 'openai', 'cohere', 'voyage', 'nvidia', 'gemini', 'mistral'
+ * @param provider - One of: 'openai', 'cohere', 'voyage', 'nvidia', 'gemini', 'mistral', 'ollama'
  * @param config - Provider-specific configuration
  * @returns EmbeddingProvider instance
  *
@@ -324,6 +384,8 @@ export declare class MistralEmbeddings implements EmbeddingProvider {
  * const embedder = getEmbedder('cohere', { model: 'embed-multilingual-v3.0' });
  * const embedder = getEmbedder('gemini'); // Uses GOOGLE_API_KEY or GEMINI_API_KEY
  * const embedder = getEmbedder('mistral'); // Uses MISTRAL_API_KEY
+ * const embedder = getEmbedder('ollama'); // Uses local Ollama server
+ * const embedder = getEmbedder('ollama', { model: 'nomic-embed-text', baseUrl: 'http://gpu:11434' });
  * ```
  */
-export declare function getEmbedder(provider: 'openai' | 'cohere' | 'voyage' | 'nvidia' | 'gemini' | 'mistral', config?: Record<string, unknown>): EmbeddingProvider;
+export declare function getEmbedder(provider: 'openai' | 'cohere' | 'voyage' | 'nvidia' | 'gemini' | 'mistral' | 'ollama', config?: Record<string, unknown>): EmbeddingProvider;

package/dist/embeddings.js

@@ -26,7 +26,7 @@
  * ```
  */
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.MistralEmbeddings = exports.GeminiEmbeddings = exports.NvidiaEmbeddings = exports.VoyageEmbeddings = exports.CohereEmbeddings = exports.OpenAIEmbeddings = exports.LOCAL_EMBEDDING_MODELS = exports.MODEL_DIMENSIONS = void 0;
+exports.OllamaEmbeddings = exports.MistralEmbeddings = exports.GeminiEmbeddings = exports.NvidiaEmbeddings = exports.VoyageEmbeddings = exports.CohereEmbeddings = exports.OpenAIEmbeddings = exports.LOCAL_EMBEDDING_MODELS = exports.MODEL_DIMENSIONS = void 0;
 exports.getEmbedder = getEmbedder;
 /**
  * Model dimension mappings for common embedding models.
@@ -575,10 +575,164 @@ class MistralEmbeddings {
     }
 }
 exports.MistralEmbeddings = MistralEmbeddings;
+/**
+ * Ollama embedding provider.
+ *
+ * Uses a local Ollama server to generate embeddings. Supports any embedding model
+ * available in Ollama, including nomic-embed-text, mxbai-embed-large, all-minilm, etc.
+ *
+ * @example
+ * ```typescript
+ * // Default: localhost:11434 with nomic-embed-text
+ * const embedder = new OllamaEmbeddings();
+ *
+ * // Custom configuration
+ * const embedder = new OllamaEmbeddings({
+ *   baseUrl: 'http://gpu-server:11434',
+ *   model: 'nomic-embed-text',
+ * });
+ *
+ * // Use with Memvid
+ * const texts = ['Hello world', 'Goodbye world'];
+ * const embeddings = await embedder.embedDocuments(texts);
+ *
+ * // Or embed and store in one step
+ * const embedding = await embedder.embedQuery('Document text...');
+ * await mem.put({
+ *   title: 'My Doc',
+ *   label: 'docs',
+ *   text: 'Document text...',
+ *   embedding,
+ *   embeddingIdentity: {
+ *     provider: 'ollama',
+ *     model: 'nomic-embed-text',
+ *     dimension: embedding.length,
+ *   },
+ * });
+ * ```
+ */
+class OllamaEmbeddings {
+    constructor(config = {}) {
+        const defaultHost = process.env.OLLAMA_HOST || 'http://localhost:11434';
+        this._baseUrl = (config.baseUrl || defaultHost).trim().replace(/\/+$/, '');
+        this._model = config.model || 'nomic-embed-text';
+        this._dimension = config.dimension;
+    }
+    get dimension() {
+        if (this._dimension)
+            return this._dimension;
+        return OllamaEmbeddings.OLLAMA_MODEL_DIMENSIONS[this._model] || 768;
+    }
+    get modelName() {
+        return this._model;
+    }
+    get provider() {
+        return 'ollama';
+    }
+    setDimensionFromEmbedding(embedding) {
+        if (!this._dimension && embedding.length > 0) {
+            this._dimension = embedding.length;
+        }
+    }
+    async embedDocuments(texts) {
+        if (texts.length === 0) {
+            return [];
+        }
+        // Ollama doesn't support batch embedding, so we process one at a time
+        // For better performance, consider using Promise.all with concurrency limit
+        const embeddings = [];
+        for (const text of texts) {
+            const response = await fetch(`${this._baseUrl}/api/embeddings`, {
+                method: 'POST',
+                headers: {
+                    'Content-Type': 'application/json',
+                },
+                body: JSON.stringify({
+                    model: this._model,
+                    prompt: text,
+                }),
+            });
+            if (!response.ok) {
+                const error = await response.text();
+                throw new Error(`Ollama API error: ${response.status} ${error}`);
+            }
+            const data = await response.json();
+            if (!Array.isArray(data.embedding)) {
+                throw new Error(`Ollama API error: invalid response format`);
+            }
+            this.setDimensionFromEmbedding(data.embedding);
+            embeddings.push(data.embedding);
+        }
+        return embeddings;
+    }
+    async embedQuery(text) {
+        const response = await fetch(`${this._baseUrl}/api/embeddings`, {
+            method: 'POST',
+            headers: {
+                'Content-Type': 'application/json',
+            },
+            body: JSON.stringify({
+                model: this._model,
+                prompt: text,
+            }),
+        });
+        if (!response.ok) {
+            const error = await response.text();
+            throw new Error(`Ollama API error: ${response.status} ${error}`);
+        }
+        const data = await response.json();
+        if (!Array.isArray(data.embedding)) {
+            throw new Error(`Ollama API error: invalid response format`);
+        }
+        this.setDimensionFromEmbedding(data.embedding);
+        return data.embedding;
+    }
+}
+exports.OllamaEmbeddings = OllamaEmbeddings;
+// Known model dimensions for popular Ollama embedding models
+OllamaEmbeddings.OLLAMA_MODEL_DIMENSIONS = {
+    // General purpose
+    'nomic-embed-text': 768,
+    'nomic-embed-text:v1': 768,
+    'nomic-embed-text:v1.5': 768,
+    'mxbai-embed-large': 1024,
+    'mxbai-embed-large:v1': 1024,
+    'all-minilm': 384,
+    'all-minilm:l6-v2': 384,
+    'all-minilm:l12-v2': 384,
+    // Snowflake Arctic (various sizes)
+    'snowflake-arctic-embed': 1024,
+    'snowflake-arctic-embed:s': 384,
+    'snowflake-arctic-embed:m': 768,
+    'snowflake-arctic-embed:l': 1024,
+    'snowflake-arctic-embed:335m': 1024,
+    // BGE models
+    'bge-m3': 1024,
+    'bge-large': 1024,
+    'bge-large:en': 1024,
+    'bge-large:en-v1.5': 1024,
+    'bge-base': 768,
+    'bge-base:en': 768,
+    'bge-base:en-v1.5': 768,
+    'bge-small': 384,
+    'bge-small:en': 384,
+    'bge-small:en-v1.5': 384,
+    // Jina embeddings
+    'jina-embeddings-v2-base-en': 768,
+    'jina-embeddings-v2-small-en': 512,
+    // Multilingual
+    'paraphrase-multilingual': 768,
+    'paraphrase-multilingual:mpnet-base-v2': 768,
+    // E5 models
+    'e5-large': 1024,
+    'e5-base': 768,
+    'e5-small': 384,
+    'e5-mistral-7b-instruct': 4096,
+};
 /**
  * Factory function to create an embedding provider.
  *
- * @param provider - One of: 'openai', 'cohere', 'voyage', 'nvidia', 'gemini', 'mistral'
+ * @param provider - One of: 'openai', 'cohere', 'voyage', 'nvidia', 'gemini', 'mistral', 'ollama'
  * @param config - Provider-specific configuration
  * @returns EmbeddingProvider instance
  *
@@ -588,6 +742,8 @@ exports.MistralEmbeddings = MistralEmbeddings;
  * const embedder = getEmbedder('cohere', { model: 'embed-multilingual-v3.0' });
  * const embedder = getEmbedder('gemini'); // Uses GOOGLE_API_KEY or GEMINI_API_KEY
  * const embedder = getEmbedder('mistral'); // Uses MISTRAL_API_KEY
+ * const embedder = getEmbedder('ollama'); // Uses local Ollama server
+ * const embedder = getEmbedder('ollama', { model: 'nomic-embed-text', baseUrl: 'http://gpu:11434' });
  * ```
  */
 function getEmbedder(provider, config) {
@@ -605,7 +761,9 @@ function getEmbedder(provider, config) {
             return new GeminiEmbeddings(config);
         case 'mistral':
             return new MistralEmbeddings(config);
+        case 'ollama':
+            return new OllamaEmbeddings(config);
         default:
-            throw new Error(`Unknown provider: ${provider}. Supported: openai, cohere, voyage, nvidia, gemini, mistral`);
+            throw new Error(`Unknown provider: ${provider}. Supported: openai, cohere, voyage, nvidia, gemini, mistral, ollama`);
     }
 }

package/dist/index.d.ts

@@ -409,7 +409,7 @@ export declare function verifyMemvid(path: string, options?: UseVerifyOptions):
 export declare function doctorMemvid(path: string, options?: UseDoctorOptions): Promise<unknown>;
 export type { AddMemoryCardsResult, Kind, ApiKey, Memvid, MemoryCard, MemoryCardInput, MemoriesResult, MemoriesStats, LockOptions, UseOptions, UnlockOptions, FindInput, VecSearchInput, AskInput, TimelineInput, PutInput, PutManyInput, PutManyOptions, MemvidErrorCode, MemvidErrorDetails, HeatmapEntry, HeatmapResponse, SessionSummary, SessionReplayResult, SessionActionResult, StatsResult, FindHit, FindResult, VecSearchResult, AskResult, AskStats, AskUsage, AskSource, Grounding, FollowUp, TimelineEntry, } from "./types";
 export { MemvidError, CapacityExceededError, TicketInvalidError, TicketReplayError, LexIndexDisabledError, TimeIndexMissingError, VerificationFailedError, LockedError, ApiKeyRequiredError, FileNotFoundError, MemoryAlreadyBoundError, FrameNotFoundError, VecIndexDisabledError, CorruptFileError, VecDimensionMismatchError, EmbeddingFailedError, EncryptionError, QuotaExceededError, getErrorSuggestion, } from "./error";
-export { EmbeddingProvider, OpenAIEmbeddings, OpenAIEmbeddingsConfig, CohereEmbeddings, CohereEmbeddingsConfig, VoyageEmbeddings, VoyageEmbeddingsConfig, NvidiaEmbeddings, NvidiaEmbeddingsConfig, GeminiEmbeddings, GeminiEmbeddingsConfig, MistralEmbeddings, MistralEmbeddingsConfig, getEmbedder, MODEL_DIMENSIONS, LOCAL_EMBEDDING_MODELS, LocalEmbeddingModel, } from "./embeddings";
+export { EmbeddingProvider, OpenAIEmbeddings, OpenAIEmbeddingsConfig, CohereEmbeddings, CohereEmbeddingsConfig, VoyageEmbeddings, VoyageEmbeddingsConfig, NvidiaEmbeddings, NvidiaEmbeddingsConfig, GeminiEmbeddings, GeminiEmbeddingsConfig, MistralEmbeddings, MistralEmbeddingsConfig, OllamaEmbeddings, OllamaEmbeddingsConfig, getEmbedder, MODEL_DIMENSIONS, LOCAL_EMBEDDING_MODELS, LocalEmbeddingModel, } from "./embeddings";
 export { flush as flushAnalytics, isTelemetryEnabled } from "./analytics";
 /**
  * Mask PII (Personally Identifiable Information) in text.

package/dist/index.js

@@ -36,7 +36,7 @@ var __exportStar = (this && this.__exportStar) || function(m, exports) {
     for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.isTelemetryEnabled = exports.flushAnalytics = exports.LOCAL_EMBEDDING_MODELS = exports.MODEL_DIMENSIONS = exports.getEmbedder = exports.MistralEmbeddings = exports.GeminiEmbeddings = exports.NvidiaEmbeddings = exports.VoyageEmbeddings = exports.CohereEmbeddings = exports.OpenAIEmbeddings = exports.getErrorSuggestion = exports.QuotaExceededError = exports.EncryptionError = exports.EmbeddingFailedError = exports.VecDimensionMismatchError = exports.CorruptFileError = exports.VecIndexDisabledError = exports.FrameNotFoundError = exports.MemoryAlreadyBoundError = exports.FileNotFoundError = exports.ApiKeyRequiredError = exports.LockedError = exports.VerificationFailedError = exports.TimeIndexMissingError = exports.LexIndexDisabledError = exports.TicketReplayError = exports.TicketInvalidError = exports.CapacityExceededError = exports.MemvidError = exports.use = exports.GeminiEntities = exports.ClaudeEntities = exports.OpenAIEntities = exports.LocalNER = exports.getEntityExtractor = exports.GeminiClip = exports.OpenAIClip = exports.LocalClip = exports.getClipProvider = exports.entities = exports.clip = void 0;
+exports.isTelemetryEnabled = exports.flushAnalytics = exports.LOCAL_EMBEDDING_MODELS = exports.MODEL_DIMENSIONS = exports.getEmbedder = exports.OllamaEmbeddings = exports.MistralEmbeddings = exports.GeminiEmbeddings = exports.NvidiaEmbeddings = exports.VoyageEmbeddings = exports.CohereEmbeddings = exports.OpenAIEmbeddings = exports.getErrorSuggestion = exports.QuotaExceededError = exports.EncryptionError = exports.EmbeddingFailedError = exports.VecDimensionMismatchError = exports.CorruptFileError = exports.VecIndexDisabledError = exports.FrameNotFoundError = exports.MemoryAlreadyBoundError = exports.FileNotFoundError = exports.ApiKeyRequiredError = exports.LockedError = exports.VerificationFailedError = exports.TimeIndexMissingError = exports.LexIndexDisabledError = exports.TicketReplayError = exports.TicketInvalidError = exports.CapacityExceededError = exports.MemvidError = exports.use = exports.GeminiEntities = exports.ClaudeEntities = exports.OpenAIEntities = exports.LocalNER = exports.getEntityExtractor = exports.GeminiClip = exports.OpenAIClip = exports.LocalClip = exports.getClipProvider = exports.entities = exports.clip = void 0;
 exports.configure = configure;
 exports.getConfig = getConfig;
 exports.resetConfig = resetConfig;
@@ -829,6 +829,8 @@ function normalisePutArgs(input) {
         extractDates: input.extractDates,
         vectorCompression: input.vectorCompression,
         timestamp: input.timestamp,
+        embedding: input.embedding,
+        embeddingIdentity: input.embeddingIdentity,
     };
     return payload;
 }
@@ -2268,6 +2270,7 @@ Object.defineProperty(exports, "VoyageEmbeddings", { enumerable: true, get: func
 Object.defineProperty(exports, "NvidiaEmbeddings", { enumerable: true, get: function () { return embeddings_1.NvidiaEmbeddings; } });
 Object.defineProperty(exports, "GeminiEmbeddings", { enumerable: true, get: function () { return embeddings_1.GeminiEmbeddings; } });
 Object.defineProperty(exports, "MistralEmbeddings", { enumerable: true, get: function () { return embeddings_1.MistralEmbeddings; } });
+Object.defineProperty(exports, "OllamaEmbeddings", { enumerable: true, get: function () { return embeddings_1.OllamaEmbeddings; } });
 Object.defineProperty(exports, "getEmbedder", { enumerable: true, get: function () { return embeddings_1.getEmbedder; } });
 Object.defineProperty(exports, "MODEL_DIMENSIONS", { enumerable: true, get: function () { return embeddings_1.MODEL_DIMENSIONS; } });
 Object.defineProperty(exports, "LOCAL_EMBEDDING_MODELS", { enumerable: true, get: function () { return embeddings_1.LOCAL_EMBEDDING_MODELS; } });

package/dist/types.d.ts

@@ -43,6 +43,10 @@ export interface PutInput {
      * "Jan 15, 2023", "2023-01-15", "01/15/2023"
      */
     timestamp?: number | string;
+    /** Optional pre-computed embedding vector (use with external embedding providers like Ollama) */
+    embedding?: number[];
+    /** Optional embedding identity metadata for `embedding` (enables CLI/SDK auto-detection). */
+    embeddingIdentity?: EmbeddingIdentity;
 }
 export interface FindInput {
     /**
@@ -211,6 +215,10 @@ export interface NativePutArgs {
     vectorCompression?: boolean;
     /** Timestamp (epoch seconds or human-readable string) */
     timestamp?: number | string;
+    /** Optional pre-computed embedding vector */
+    embedding?: number[];
+    /** Optional embedding identity metadata */
+    embeddingIdentity?: EmbeddingIdentity;
 }
 export interface NativeFindOptions {
     k?: number;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@memvid/sdk",
-  "version": "2.0.152",
+  "version": "2.0.153",
   "description": "Single-file AI memory system for Node.js. Store, search, and query documents with built-in RAG.",
   "main": "./dist/index.js",
   "types": "./dist/index.d.ts",
@@ -41,11 +41,11 @@
     "node": ">=18"
   },
   "optionalDependencies": {
-    "@memvid/sdk-darwin-arm64": "2.0.152",
-    "@memvid/sdk-darwin-x64": "2.0.152",
-    "@memvid/sdk-linux-x64-gnu": "2.0.152",
-    "@memvid/sdk-linux-arm64-gnu": "2.0.152",
-    "@memvid/sdk-win32-x64-msvc": "2.0.152"
+    "@memvid/sdk-darwin-arm64": "2.0.153",
+    "@memvid/sdk-darwin-x64": "2.0.153",
+    "@memvid/sdk-linux-x64-gnu": "2.0.153",
+    "@memvid/sdk-linux-arm64-gnu": "2.0.153",
+    "@memvid/sdk-win32-x64-msvc": "2.0.153"
   },
   "peerDependencies": {
     "@langchain/core": ">=0.3.0",