@omendb/omendb 0.0.9 → 0.0.11

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/index.d.ts +249 -126
  2. package/package.json +5 -5
package/index.d.ts CHANGED
@@ -1,103 +1,200 @@
1
1
  /* auto-generated by NAPI-RS */
2
2
  /* eslint-disable */
3
3
  export declare class VectorDatabase {
4
- /**
5
- * Insert or update vectors.
6
- *
7
- * Accepts an array of items with id, vector, and optional metadata.
8
- */
9
- set(items: Array<VectorItem>): Array<number>
10
- /**
11
- * Search for k nearest neighbors.
12
- *
13
- * @param query - Query vector (number[] or Float32Array)
14
- * @param k - Number of results to return
15
- * @param ef - Optional search width override
16
- * @param filter - Optional metadata filter (e.g., {category: "foo"} or {price: {$gt: 10}})
17
- * @returns Array of {id, distance, metadata}
18
- */
19
- search(query: Array<number> | Float32Array, k: number, ef?: number | undefined | null, filter?: Record<string, unknown> | undefined): Array<SearchResult>
20
- /**
21
- * Batch search with parallel execution (async).
22
- *
23
- * Runs searches in parallel using rayon, returns Promise.
24
- */
25
- searchBatch(queries: Array<Array<number> | Float32Array>, k: number, ef?: number | undefined | null): Promise<Array<Array<SearchResult>>>
26
- /** Get a vector by ID. */
27
- get(id: string): GetResult | null
28
- /**
29
- * Delete vectors by ID.
30
- *
31
- * @returns Number of vectors deleted
32
- */
33
- delete(ids: Array<string>): number
34
- /** Update a vector's data and/or metadata. */
35
- update(id: string, vector: Array<number> | Float32Array, metadata?: Record<string, unknown> | undefined): void
36
- /** Save database to disk. */
37
- save(): void
38
- /** Get number of vectors in database. */
39
- get count(): number
40
- /** Get current ef_search value. */
41
- get efSearch(): number | null
42
- /** Set ef_search value. */
43
- set efSearch(efSearch: number)
44
- /** Get or create a named collection. */
45
- collection(name: string): VectorDatabase
46
- /** List all collections. */
47
- collections(): Array<string>
48
- /** Delete a collection. */
49
- deleteCollection(name: string): void
50
- /**
51
- * Enable text search for hybrid (vector + text) search.
52
- *
53
- * Must be called before using setWithText() or hybridSearch().
54
- */
55
- enableTextSearch(): void
56
- /** Check if text search is enabled. */
57
- get hasTextSearch(): boolean
58
- /**
59
- * Set vectors with associated text for hybrid search.
60
- *
61
- * @param items - Array of {id, vector, text, metadata?}
62
- * @returns Array of internal indices
63
- */
64
- setWithText(items: Array<VectorItemWithText>): Array<number>
65
- /**
66
- * Search using text only (BM25 scoring).
67
- *
68
- * @param query - Text query
69
- * @param k - Number of results
70
- * @returns Array of {id, score, metadata}
71
- */
72
- textSearch(query: string, k: number): Array<TextSearchResult>
73
- /**
74
- * Hybrid search combining vector similarity and text relevance.
75
- *
76
- * Uses Reciprocal Rank Fusion (RRF) to combine HNSW and BM25 results.
77
- *
78
- * @param queryVector - Query embedding
79
- * @param queryText - Text query for BM25
80
- * @param k - Number of results
81
- * @param filter - Optional metadata filter
82
- * @param alpha - Weight for vector vs text (0.0=text only, 1.0=vector only, default=0.5)
83
- * @param rrfK - RRF constant (default=60, higher reduces rank influence)
84
- * @returns Array of {id, score, metadata}
85
- */
86
- hybridSearch(queryVector: Array<number> | Float32Array, queryText: string, k: number, filter?: Record<string, unknown> | undefined, alpha?: number | undefined | null, rrfK?: number | undefined | null): Array<TextSearchResult>
87
- /**
88
- * Flush pending changes to disk.
89
- *
90
- * For hybrid search, this commits text index changes.
91
- */
92
- flush(): void
93
- /** Merge another database into this one. */
94
- mergeFrom(other: VectorDatabase): number
4
+ /**
5
+ * Insert or update vectors.
6
+ *
7
+ * Accepts an array of items with id, vector, and optional metadata.
8
+ */
9
+ set(items: Array<VectorItem>): Array<number>;
10
+ /**
11
+ * Search for k nearest neighbors.
12
+ *
13
+ * @param query - Query vector (number[] or Float32Array)
14
+ * @param k - Number of results to return
15
+ * @param ef - Optional search width override
16
+ * @param filter - Optional metadata filter (e.g., {category: "foo"} or {price: {$gt: 10}})
17
+ * @returns Array of {id, distance, metadata}
18
+ */
19
+ search(
20
+ query: Array<number> | Float32Array,
21
+ k: number,
22
+ ef?: number | undefined | null,
23
+ filter?: Record<string, unknown> | undefined,
24
+ ): Array<SearchResult>;
25
+ /**
26
+ * Batch search with parallel execution (async).
27
+ *
28
+ * Runs searches in parallel using rayon, returns Promise.
29
+ */
30
+ searchBatch(
31
+ queries: Array<Array<number> | Float32Array>,
32
+ k: number,
33
+ ef?: number | undefined | null,
34
+ ): Promise<Array<Array<SearchResult>>>;
35
+ /** Get a vector by ID. */
36
+ get(id: string): GetResult | null;
37
+ /**
38
+ * Delete vectors by ID.
39
+ *
40
+ * @returns Number of vectors deleted
41
+ */
42
+ delete(ids: Array<string>): number;
43
+ /**
44
+ * Delete vectors matching a metadata filter.
45
+ *
46
+ * Evaluates the filter against all vectors and deletes those that match.
47
+ * Uses the same MongoDB-style filter syntax as search().
48
+ *
49
+ * @param filter - MongoDB-style metadata filter
50
+ * @returns Number of vectors deleted
51
+ *
52
+ * @example
53
+ * ```javascript
54
+ * // Delete by equality
55
+ * db.deleteWhere({ status: "archived" });
56
+ *
57
+ * // Delete with comparison
58
+ * db.deleteWhere({ score: { $lt: 0.5 } });
59
+ *
60
+ * // Complex filter
61
+ * db.deleteWhere({ $and: [{ type: "draft" }, { age: { $gt: 30 } }] });
62
+ * ```
63
+ */
64
+ deleteWhere(filter: Record<string, unknown>): number;
65
+ /**
66
+ * Count vectors, optionally filtered by metadata.
67
+ *
68
+ * Without a filter, returns total count (same as db.length).
69
+ * With a filter, returns count of vectors matching the filter.
70
+ *
71
+ * @param filter - Optional MongoDB-style metadata filter
72
+ * @returns Number of vectors (matching filter if provided)
73
+ *
74
+ * @example
75
+ * ```javascript
76
+ * // Total count
77
+ * const total = db.count();
78
+ *
79
+ * // Filtered count
80
+ * const active = db.count({ status: "active" });
81
+ *
82
+ * // With comparison operators
83
+ * const highScore = db.count({ score: { $gte: 0.8 } });
84
+ * ```
85
+ */
86
+ count(filter?: Record<string, unknown> | undefined): number;
87
+ /** Update a vector's data and/or metadata. */
88
+ update(
89
+ id: string,
90
+ vector: Array<number> | Float32Array,
91
+ metadata?: Record<string, unknown> | undefined,
92
+ ): void;
93
+ /** Get number of vectors in database. */
94
+ get length(): number;
95
+ /** Get current ef_search value. */
96
+ get efSearch(): number;
97
+ /** Set ef_search value. */
98
+ set efSearch(efSearch: number);
99
+ /**
100
+ * Get or create a named collection.
101
+ *
102
+ * Collection handles share state - changes made through one handle
103
+ * are immediately visible through another (no flush required).
104
+ */
105
+ collection(name: string): VectorDatabase;
106
+ /** List all collections. */
107
+ collections(): Array<string>;
108
+ /** Delete a collection. */
109
+ deleteCollection(name: string): void;
110
+ /**
111
+ * Enable text search for hybrid (vector + text) search.
112
+ *
113
+ * Must be called before using setWithText() or hybridSearch().
114
+ */
115
+ enableTextSearch(): void;
116
+ /** Check if text search is enabled. */
117
+ get hasTextSearch(): boolean;
118
+ /**
119
+ * Set vectors with associated text for hybrid search.
120
+ *
121
+ * @param items - Array of {id, vector, text, metadata?}
122
+ * @returns Array of internal indices
123
+ */
124
+ setWithText(items: Array<VectorItemWithText>): Array<number>;
125
+ /**
126
+ * Search using text only (BM25 scoring).
127
+ *
128
+ * @param query - Text query
129
+ * @param k - Number of results
130
+ * @returns Array of {id, score, metadata}
131
+ */
132
+ textSearch(query: string, k: number): Array<TextSearchResult>;
133
+ /**
134
+ * Hybrid search combining vector similarity and text relevance.
135
+ *
136
+ * Uses Reciprocal Rank Fusion (RRF) to combine HNSW and BM25 results.
137
+ *
138
+ * @param queryVector - Query embedding
139
+ * @param queryText - Text query for BM25
140
+ * @param k - Number of results
141
+ * @param filter - Optional metadata filter
142
+ * @param alpha - Weight for vector vs text (0.0=text only, 1.0=vector only, default=0.5)
143
+ * @param rrfK - RRF constant (default=60, higher reduces rank influence)
144
+ * @returns Array of {id, score, metadata}
145
+ */
146
+ hybridSearch(
147
+ queryVector: Array<number> | Float32Array,
148
+ queryText: string,
149
+ k: number,
150
+ filter?: Record<string, unknown> | undefined,
151
+ alpha?: number | undefined | null,
152
+ rrfK?: number | undefined | null,
153
+ ): Array<TextSearchResult>;
154
+ /**
155
+ * Flush pending changes to disk.
156
+ *
157
+ * For hybrid search, this commits text index changes.
158
+ */
159
+ flush(): void;
160
+ /** Merge another database into this one. */
161
+ mergeFrom(other: VectorDatabase): number;
162
+ /**
163
+ * List all vector IDs (without loading vector data).
164
+ *
165
+ * Efficient way to get all IDs for iteration, export, or debugging.
166
+ * @returns Array of all vector IDs in the database
167
+ */
168
+ ids(): Array<string>;
169
+ /**
170
+ * Get all items as array of {id, vector, metadata}.
171
+ *
172
+ * Returns all vectors with their IDs and metadata.
173
+ * For large datasets, consider using ids() and get() in batches.
174
+ */
175
+ items(): Array<GetResult>;
176
+ /**
177
+ * Check if an ID exists in the database.
178
+ *
179
+ * @param id - Vector ID to check
180
+ * @returns true if ID exists and is not deleted
181
+ */
182
+ exists(id: string): boolean;
183
+ /**
184
+ * Get multiple vectors by ID.
185
+ *
186
+ * Batch version of get(). More efficient than calling get() in a loop.
187
+ *
188
+ * @param ids - Array of vector IDs to retrieve
189
+ * @returns Array of results in same order as input, null for missing IDs
190
+ */
191
+ getMany(ids: Array<string>): Array<GetResult | undefined | null>;
95
192
  }
96
193
 
97
194
  export interface GetResult {
98
- id: string
99
- vector: Array<number>
100
- metadata: Record<string, unknown>
195
+ id: string;
196
+ vector: Array<number>;
197
+ metadata: Record<string, unknown>;
101
198
  }
102
199
 
103
200
  /**
@@ -125,9 +222,20 @@ export interface GetResult {
125
222
  * dimensions: 128,
126
223
  * quantization: 4 // 4-bit quantization
127
224
  * });
225
+ *
226
+ * // Quantization with custom rescore settings
227
+ * const db = omendb.open("./mydb", {
228
+ * dimensions: 128,
229
+ * quantization: 4,
230
+ * rescore: false, // Disable rescore for max speed
231
+ * oversample: 5.0 // Or increase oversample for better recall
232
+ * });
128
233
  * ```
129
234
  */
130
- export declare function open(path: string, options?: OpenOptions | undefined | null): VectorDatabase
235
+ export declare function open(
236
+ path: string,
237
+ options?: OpenOptions | undefined | null,
238
+ ): VectorDatabase;
131
239
 
132
240
  /**
133
241
  * Configuration options for opening a vector database.
@@ -138,49 +246,64 @@ export declare function open(path: string, options?: OpenOptions | undefined | n
138
246
  * - efConstruction: 100 (build quality, higher = better graph, slower build)
139
247
  * - efSearch: 100 (search quality, higher = better recall, slower search)
140
248
  * - quantization: null (RaBitQ bit width: 2, 4, or 8 for compression)
249
+ * - rescore: true when quantization enabled (rerank candidates with exact distance)
250
+ * - oversample: 3.0 (fetch k*oversample candidates when rescoring)
251
+ * - metric: "l2" (distance metric: "l2", "euclidean", "cosine", "dot", "ip")
141
252
  */
142
253
  export interface OpenOptions {
143
- /** Vector dimensions (default: 128, auto-detected on first insert) */
144
- dimensions?: number
145
- /** HNSW M parameter: neighbors per node (default: 16, range: 4-64) */
146
- m?: number
147
- /** HNSW ef_construction: build quality (default: 100, must be >= m) */
148
- efConstruction?: number
149
- /** HNSW ef_search: search quality/speed tradeoff (default: 100) */
150
- efSearch?: number
151
- /**
152
- * RaBitQ quantization bits: 2, 4, or 8 (default: null = no quantization)
153
- * Enables 4-16x memory compression with ~1-2% recall loss
154
- */
155
- quantization?: number
254
+ /** Vector dimensions (default: 128, auto-detected on first insert) */
255
+ dimensions?: number;
256
+ /** HNSW M parameter: neighbors per node (default: 16, range: 4-64) */
257
+ m?: number;
258
+ /** HNSW ef_construction: build quality (default: 100, must be >= m) */
259
+ efConstruction?: number;
260
+ /** HNSW ef_search: search quality/speed tradeoff (default: 100) */
261
+ efSearch?: number;
262
+ /**
263
+ * RaBitQ quantization bits: 2, 4, or 8 (default: null = no quantization)
264
+ * Enables 4-16x memory compression with ~1-2% recall loss
265
+ */
266
+ quantization?: number;
267
+ /**
268
+ * Rescore candidates with exact distance (default: true when quantization enabled)
269
+ * Set to false for maximum speed at the cost of ~20% recall
270
+ */
271
+ rescore?: boolean;
272
+ /**
273
+ * Oversampling factor for rescoring (default: 3.0)
274
+ * Fetches k*oversample candidates then reranks to return top k
275
+ */
276
+ oversample?: number;
277
+ /** Distance metric: "l2"/"euclidean" (default), "cosine", "dot"/"ip" */
278
+ metric?: string;
156
279
  }
157
280
 
158
281
  export interface SearchResult {
159
- id: string
160
- distance: number
161
- /** Metadata as JSON (using serde-json feature) */
162
- metadata: Record<string, unknown>
282
+ id: string;
283
+ distance: number;
284
+ /** Metadata as JSON (using serde-json feature) */
285
+ metadata: Record<string, unknown>;
163
286
  }
164
287
 
165
288
  export interface TextSearchResult {
166
- id: string
167
- score: number
168
- metadata: Record<string, unknown>
289
+ id: string;
290
+ score: number;
291
+ metadata: Record<string, unknown>;
169
292
  }
170
293
 
171
294
  export interface VectorItem {
172
- id: string
173
- /** Vector data as array of numbers */
174
- vector: Array<number>
175
- /** Optional metadata */
176
- metadata?: Record<string, unknown> | undefined
177
- /** Optional document text (stored in metadata.document) */
178
- document?: string
295
+ id: string;
296
+ /** Vector data as array of numbers */
297
+ vector: Array<number>;
298
+ /** Optional metadata */
299
+ metadata?: Record<string, unknown> | undefined;
300
+ /** Optional document text (stored in metadata.document) */
301
+ document?: string;
179
302
  }
180
303
 
181
304
  export interface VectorItemWithText {
182
- id: string
183
- vector: Array<number>
184
- text: string
185
- metadata?: Record<string, unknown> | undefined
305
+ id: string;
306
+ vector: Array<number>;
307
+ text: string;
308
+ metadata?: Record<string, unknown> | undefined;
186
309
  }
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@omendb/omendb",
3
- "version": "0.0.9",
3
+ "version": "0.0.11",
4
4
  "description": "Fast embedded vector database with HNSW indexing",
5
5
  "main": "index.js",
6
6
  "types": "index.d.ts",
@@ -49,9 +49,9 @@
49
49
  "omendb.node"
50
50
  ],
51
51
  "optionalDependencies": {
52
- "@omendb/omendb-darwin-x64": "0.0.9",
53
- "@omendb/omendb-darwin-arm64": "0.0.9",
54
- "@omendb/omendb-linux-x64-gnu": "0.0.9",
55
- "@omendb/omendb-linux-arm64-gnu": "0.0.9"
52
+ "@omendb/omendb-darwin-x64": "0.0.11",
53
+ "@omendb/omendb-darwin-arm64": "0.0.11",
54
+ "@omendb/omendb-linux-x64-gnu": "0.0.11",
55
+ "@omendb/omendb-linux-arm64-gnu": "0.0.11"
56
56
  }
57
57
  }