ruvector 0.2.20 → 0.2.22

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.
@@ -0,0 +1,220 @@
1
+ /* @ts-self-types="./ruvector_decompiler_wasm.d.ts" */
2
+
3
+ /**
4
+ * Decompile a minified JavaScript bundle using the full Louvain pipeline.
5
+ *
6
+ * # Arguments
7
+ *
8
+ * * `source` - The minified JavaScript source code.
9
+ * * `config_json` - JSON string of `DecompileConfig` fields. Pass `"{}"` for defaults.
10
+ *
11
+ * # Returns
12
+ *
13
+ * A JSON string containing the `DecompileResult` (modules, witness, inferred names, etc.)
14
+ * or a JSON object with an `"error"` field on failure.
15
+ * @param {string} source
16
+ * @param {string} config_json
17
+ * @returns {string}
18
+ */
19
+ function decompile(source, config_json) {
20
+ let deferred3_0;
21
+ let deferred3_1;
22
+ try {
23
+ const retptr = wasm.__wbindgen_add_to_stack_pointer(-16);
24
+ const ptr0 = passStringToWasm0(source, wasm.__wbindgen_export2, wasm.__wbindgen_export3);
25
+ const len0 = WASM_VECTOR_LEN;
26
+ const ptr1 = passStringToWasm0(config_json, wasm.__wbindgen_export2, wasm.__wbindgen_export3);
27
+ const len1 = WASM_VECTOR_LEN;
28
+ wasm.decompile(retptr, ptr0, len0, ptr1, len1);
29
+ var r0 = getDataViewMemory0().getInt32(retptr + 4 * 0, true);
30
+ var r1 = getDataViewMemory0().getInt32(retptr + 4 * 1, true);
31
+ deferred3_0 = r0;
32
+ deferred3_1 = r1;
33
+ return getStringFromWasm0(r0, r1);
34
+ } finally {
35
+ wasm.__wbindgen_add_to_stack_pointer(16);
36
+ wasm.__wbindgen_export(deferred3_0, deferred3_1, 1);
37
+ }
38
+ }
39
+ exports.decompile = decompile;
40
+
41
+ /**
42
+ * Initialize the WASM module (sets up panic hook for better error messages).
43
+ */
44
+ function init() {
45
+ wasm.init();
46
+ }
47
+ exports.init = init;
48
+
49
+ /**
50
+ * Return the version of the decompiler WASM module.
51
+ * @returns {string}
52
+ */
53
+ function version() {
54
+ let deferred1_0;
55
+ let deferred1_1;
56
+ try {
57
+ const retptr = wasm.__wbindgen_add_to_stack_pointer(-16);
58
+ wasm.version(retptr);
59
+ var r0 = getDataViewMemory0().getInt32(retptr + 4 * 0, true);
60
+ var r1 = getDataViewMemory0().getInt32(retptr + 4 * 1, true);
61
+ deferred1_0 = r0;
62
+ deferred1_1 = r1;
63
+ return getStringFromWasm0(r0, r1);
64
+ } finally {
65
+ wasm.__wbindgen_add_to_stack_pointer(16);
66
+ wasm.__wbindgen_export(deferred1_0, deferred1_1, 1);
67
+ }
68
+ }
69
+ exports.version = version;
70
+
71
+ function __wbg_get_imports() {
72
+ const import0 = {
73
+ __proto__: null,
74
+ __wbg___wbindgen_throw_39bc967c0e5a9b58: function(arg0, arg1) {
75
+ throw new Error(getStringFromWasm0(arg0, arg1));
76
+ },
77
+ __wbg_error_a6fa202b58aa1cd3: function(arg0, arg1) {
78
+ let deferred0_0;
79
+ let deferred0_1;
80
+ try {
81
+ deferred0_0 = arg0;
82
+ deferred0_1 = arg1;
83
+ console.error(getStringFromWasm0(arg0, arg1));
84
+ } finally {
85
+ wasm.__wbindgen_export(deferred0_0, deferred0_1, 1);
86
+ }
87
+ },
88
+ __wbg_new_227d7c05414eb861: function() {
89
+ const ret = new Error();
90
+ return addHeapObject(ret);
91
+ },
92
+ __wbg_stack_3b0d974bbf31e44f: function(arg0, arg1) {
93
+ const ret = getObject(arg1).stack;
94
+ const ptr1 = passStringToWasm0(ret, wasm.__wbindgen_export2, wasm.__wbindgen_export3);
95
+ const len1 = WASM_VECTOR_LEN;
96
+ getDataViewMemory0().setInt32(arg0 + 4 * 1, len1, true);
97
+ getDataViewMemory0().setInt32(arg0 + 4 * 0, ptr1, true);
98
+ },
99
+ __wbindgen_object_drop_ref: function(arg0) {
100
+ takeObject(arg0);
101
+ },
102
+ };
103
+ return {
104
+ __proto__: null,
105
+ "./ruvector_decompiler_wasm_bg.js": import0,
106
+ };
107
+ }
108
+
109
+ function addHeapObject(obj) {
110
+ if (heap_next === heap.length) heap.push(heap.length + 1);
111
+ const idx = heap_next;
112
+ heap_next = heap[idx];
113
+
114
+ heap[idx] = obj;
115
+ return idx;
116
+ }
117
+
118
+ function dropObject(idx) {
119
+ if (idx < 1028) return;
120
+ heap[idx] = heap_next;
121
+ heap_next = idx;
122
+ }
123
+
124
+ let cachedDataViewMemory0 = null;
125
+ function getDataViewMemory0() {
126
+ if (cachedDataViewMemory0 === null || cachedDataViewMemory0.buffer.detached === true || (cachedDataViewMemory0.buffer.detached === undefined && cachedDataViewMemory0.buffer !== wasm.memory.buffer)) {
127
+ cachedDataViewMemory0 = new DataView(wasm.memory.buffer);
128
+ }
129
+ return cachedDataViewMemory0;
130
+ }
131
+
132
+ function getStringFromWasm0(ptr, len) {
133
+ ptr = ptr >>> 0;
134
+ return decodeText(ptr, len);
135
+ }
136
+
137
+ let cachedUint8ArrayMemory0 = null;
138
+ function getUint8ArrayMemory0() {
139
+ if (cachedUint8ArrayMemory0 === null || cachedUint8ArrayMemory0.byteLength === 0) {
140
+ cachedUint8ArrayMemory0 = new Uint8Array(wasm.memory.buffer);
141
+ }
142
+ return cachedUint8ArrayMemory0;
143
+ }
144
+
145
+ function getObject(idx) { return heap[idx]; }
146
+
147
+ let heap = new Array(1024).fill(undefined);
148
+ heap.push(undefined, null, true, false);
149
+
150
+ let heap_next = heap.length;
151
+
152
+ function passStringToWasm0(arg, malloc, realloc) {
153
+ if (realloc === undefined) {
154
+ const buf = cachedTextEncoder.encode(arg);
155
+ const ptr = malloc(buf.length, 1) >>> 0;
156
+ getUint8ArrayMemory0().subarray(ptr, ptr + buf.length).set(buf);
157
+ WASM_VECTOR_LEN = buf.length;
158
+ return ptr;
159
+ }
160
+
161
+ let len = arg.length;
162
+ let ptr = malloc(len, 1) >>> 0;
163
+
164
+ const mem = getUint8ArrayMemory0();
165
+
166
+ let offset = 0;
167
+
168
+ for (; offset < len; offset++) {
169
+ const code = arg.charCodeAt(offset);
170
+ if (code > 0x7F) break;
171
+ mem[ptr + offset] = code;
172
+ }
173
+ if (offset !== len) {
174
+ if (offset !== 0) {
175
+ arg = arg.slice(offset);
176
+ }
177
+ ptr = realloc(ptr, len, len = offset + arg.length * 3, 1) >>> 0;
178
+ const view = getUint8ArrayMemory0().subarray(ptr + offset, ptr + len);
179
+ const ret = cachedTextEncoder.encodeInto(arg, view);
180
+
181
+ offset += ret.written;
182
+ ptr = realloc(ptr, len, offset, 1) >>> 0;
183
+ }
184
+
185
+ WASM_VECTOR_LEN = offset;
186
+ return ptr;
187
+ }
188
+
189
+ function takeObject(idx) {
190
+ const ret = getObject(idx);
191
+ dropObject(idx);
192
+ return ret;
193
+ }
194
+
195
+ let cachedTextDecoder = new TextDecoder('utf-8', { ignoreBOM: true, fatal: true });
196
+ cachedTextDecoder.decode();
197
+ function decodeText(ptr, len) {
198
+ return cachedTextDecoder.decode(getUint8ArrayMemory0().subarray(ptr, ptr + len));
199
+ }
200
+
201
+ const cachedTextEncoder = new TextEncoder();
202
+
203
+ if (!('encodeInto' in cachedTextEncoder)) {
204
+ cachedTextEncoder.encodeInto = function (arg, view) {
205
+ const buf = cachedTextEncoder.encode(arg);
206
+ view.set(buf);
207
+ return {
208
+ read: arg.length,
209
+ written: buf.length
210
+ };
211
+ };
212
+ }
213
+
214
+ let WASM_VECTOR_LEN = 0;
215
+
216
+ const wasmPath = `${__dirname}/ruvector_decompiler_wasm_bg.wasm`;
217
+ const wasmBytes = require('fs').readFileSync(wasmPath);
218
+ const wasmModule = new WebAssembly.Module(wasmBytes);
219
+ let wasm = new WebAssembly.Instance(wasmModule, __wbg_get_imports()).exports;
220
+ wasm.__wbindgen_start();
@@ -0,0 +1,16 @@
1
+ /* tslint:disable */
2
+ /* eslint-disable */
3
+ export const memory: WebAssembly.Memory;
4
+ export const decompile: (a: number, b: number, c: number, d: number, e: number) => void;
5
+ export const version: (a: number) => void;
6
+ export const init: () => void;
7
+ export const mincut_add_result: (a: number, b: number, c: number, d: number, e: number, f: number, g: number) => void;
8
+ export const mincut_get_coordinator: () => number;
9
+ export const mincut_get_result: () => number;
10
+ export const mincut_init: (a: number, b: number, c: number) => void;
11
+ export const mincut_is_complete: () => number;
12
+ export const __wbindgen_export: (a: number, b: number, c: number) => void;
13
+ export const __wbindgen_export2: (a: number, b: number) => number;
14
+ export const __wbindgen_export3: (a: number, b: number, c: number, d: number) => number;
15
+ export const __wbindgen_add_to_stack_pointer: (a: number) => number;
16
+ export const __wbindgen_start: () => void;
@@ -1,348 +0,0 @@
1
- /**
2
- * Model Loader for RuVector ONNX Embeddings WASM
3
- *
4
- * Provides easy loading of pre-trained models from HuggingFace Hub
5
- */
6
-
7
- /**
8
- * Pre-configured models with their HuggingFace URLs
9
- */
10
- export const MODELS = {
11
- // Sentence Transformers - Small & Fast
12
- 'all-MiniLM-L6-v2': {
13
- name: 'all-MiniLM-L6-v2',
14
- dimension: 384,
15
- maxLength: 256,
16
- size: '23MB',
17
- description: 'Fast, general-purpose embeddings',
18
- model: 'https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2/resolve/main/onnx/model.onnx',
19
- tokenizer: 'https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2/resolve/main/tokenizer.json',
20
- },
21
- 'all-MiniLM-L12-v2': {
22
- name: 'all-MiniLM-L12-v2',
23
- dimension: 384,
24
- maxLength: 256,
25
- size: '33MB',
26
- description: 'Better quality, balanced speed',
27
- model: 'https://huggingface.co/sentence-transformers/all-MiniLM-L12-v2/resolve/main/onnx/model.onnx',
28
- tokenizer: 'https://huggingface.co/sentence-transformers/all-MiniLM-L12-v2/resolve/main/tokenizer.json',
29
- },
30
-
31
- // BGE Models - State of the art
32
- 'bge-small-en-v1.5': {
33
- name: 'bge-small-en-v1.5',
34
- dimension: 384,
35
- maxLength: 512,
36
- size: '33MB',
37
- description: 'State-of-the-art small model',
38
- model: 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/onnx/model.onnx',
39
- tokenizer: 'https://huggingface.co/BAAI/bge-small-en-v1.5/resolve/main/tokenizer.json',
40
- },
41
- 'bge-base-en-v1.5': {
42
- name: 'bge-base-en-v1.5',
43
- dimension: 768,
44
- maxLength: 512,
45
- size: '110MB',
46
- description: 'Best overall quality',
47
- model: 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/onnx/model.onnx',
48
- tokenizer: 'https://huggingface.co/BAAI/bge-base-en-v1.5/resolve/main/tokenizer.json',
49
- },
50
-
51
- // E5 Models - Microsoft
52
- 'e5-small-v2': {
53
- name: 'e5-small-v2',
54
- dimension: 384,
55
- maxLength: 512,
56
- size: '33MB',
57
- description: 'Excellent for search & retrieval',
58
- model: 'https://huggingface.co/intfloat/e5-small-v2/resolve/main/onnx/model.onnx',
59
- tokenizer: 'https://huggingface.co/intfloat/e5-small-v2/resolve/main/tokenizer.json',
60
- },
61
-
62
- // GTE Models - Alibaba
63
- 'gte-small': {
64
- name: 'gte-small',
65
- dimension: 384,
66
- maxLength: 512,
67
- size: '33MB',
68
- description: 'Good multilingual support',
69
- model: 'https://huggingface.co/thenlper/gte-small/resolve/main/onnx/model.onnx',
70
- tokenizer: 'https://huggingface.co/thenlper/gte-small/resolve/main/tokenizer.json',
71
- },
72
- };
73
-
74
- /**
75
- * Default model for quick start
76
- */
77
- export const DEFAULT_MODEL = 'all-MiniLM-L6-v2';
78
-
79
- /**
80
- * Model loader with caching support
81
- */
82
- export class ModelLoader {
83
- constructor(options = {}) {
84
- this.cache = options.cache ?? true;
85
- this.cacheStorage = options.cacheStorage ?? 'ruvector-models';
86
- this.onProgress = options.onProgress ?? null;
87
- }
88
-
89
- /**
90
- * Load a pre-configured model by name
91
- * @param {string} modelName - Model name from MODELS
92
- * @returns {Promise<{modelBytes: Uint8Array, tokenizerJson: string, config: object}>}
93
- */
94
- async loadModel(modelName = DEFAULT_MODEL) {
95
- const modelConfig = MODELS[modelName];
96
- if (!modelConfig) {
97
- throw new Error(`Unknown model: ${modelName}. Available: ${Object.keys(MODELS).join(', ')}`);
98
- }
99
-
100
- console.log(`Loading model: ${modelConfig.name} (${modelConfig.size})`);
101
-
102
- const [modelBytes, tokenizerJson] = await Promise.all([
103
- this.fetchWithCache(modelConfig.model, `${modelName}-model.onnx`, 'arraybuffer'),
104
- this.fetchWithCache(modelConfig.tokenizer, `${modelName}-tokenizer.json`, 'text'),
105
- ]);
106
-
107
- return {
108
- modelBytes: new Uint8Array(modelBytes),
109
- tokenizerJson,
110
- config: modelConfig,
111
- };
112
- }
113
-
114
- /**
115
- * Load model from custom URLs
116
- * @param {string} modelUrl - URL to ONNX model
117
- * @param {string} tokenizerUrl - URL to tokenizer.json
118
- * @returns {Promise<{modelBytes: Uint8Array, tokenizerJson: string}>}
119
- */
120
- async loadFromUrls(modelUrl, tokenizerUrl) {
121
- const [modelBytes, tokenizerJson] = await Promise.all([
122
- this.fetchWithCache(modelUrl, null, 'arraybuffer'),
123
- this.fetchWithCache(tokenizerUrl, null, 'text'),
124
- ]);
125
-
126
- return {
127
- modelBytes: new Uint8Array(modelBytes),
128
- tokenizerJson,
129
- };
130
- }
131
-
132
- /**
133
- * Load model from local files (Node.js)
134
- * @param {string} modelPath - Path to ONNX model
135
- * @param {string} tokenizerPath - Path to tokenizer.json
136
- * @returns {Promise<{modelBytes: Uint8Array, tokenizerJson: string}>}
137
- */
138
- async loadFromFiles(modelPath, tokenizerPath) {
139
- // Node.js environment
140
- if (typeof process !== 'undefined' && process.versions?.node) {
141
- const fs = await import('fs/promises');
142
- const [modelBytes, tokenizerJson] = await Promise.all([
143
- fs.readFile(modelPath),
144
- fs.readFile(tokenizerPath, 'utf8'),
145
- ]);
146
- return {
147
- modelBytes: new Uint8Array(modelBytes),
148
- tokenizerJson,
149
- };
150
- }
151
- throw new Error('loadFromFiles is only available in Node.js');
152
- }
153
-
154
- /**
155
- * Fetch with optional caching (uses Cache API in browsers)
156
- */
157
- async fetchWithCache(url, cacheKey, responseType) {
158
- // Try cache first (browser only)
159
- if (this.cache && typeof caches !== 'undefined' && cacheKey) {
160
- try {
161
- const cache = await caches.open(this.cacheStorage);
162
- const cached = await cache.match(cacheKey);
163
- if (cached) {
164
- console.log(` Cache hit: ${cacheKey}`);
165
- return responseType === 'arraybuffer'
166
- ? await cached.arrayBuffer()
167
- : await cached.text();
168
- }
169
- } catch (e) {
170
- // Cache API not available, continue with fetch
171
- }
172
- }
173
-
174
- // Fetch from network
175
- console.log(` Downloading: ${url}`);
176
- const response = await this.fetchWithProgress(url);
177
-
178
- if (!response.ok) {
179
- throw new Error(`Failed to fetch ${url}: ${response.status} ${response.statusText}`);
180
- }
181
-
182
- // Clone for caching
183
- const responseClone = response.clone();
184
-
185
- // Cache the response (browser only)
186
- if (this.cache && typeof caches !== 'undefined' && cacheKey) {
187
- try {
188
- const cache = await caches.open(this.cacheStorage);
189
- await cache.put(cacheKey, responseClone);
190
- } catch (e) {
191
- // Cache write failed, continue
192
- }
193
- }
194
-
195
- return responseType === 'arraybuffer'
196
- ? await response.arrayBuffer()
197
- : await response.text();
198
- }
199
-
200
- /**
201
- * Fetch with progress reporting
202
- */
203
- async fetchWithProgress(url) {
204
- const response = await fetch(url);
205
-
206
- if (!this.onProgress || !response.body) {
207
- return response;
208
- }
209
-
210
- const contentLength = response.headers.get('content-length');
211
- if (!contentLength) {
212
- return response;
213
- }
214
-
215
- const total = parseInt(contentLength, 10);
216
- let loaded = 0;
217
-
218
- const reader = response.body.getReader();
219
- const chunks = [];
220
-
221
- while (true) {
222
- const { done, value } = await reader.read();
223
- if (done) break;
224
-
225
- chunks.push(value);
226
- loaded += value.length;
227
-
228
- this.onProgress({
229
- loaded,
230
- total,
231
- percent: Math.round((loaded / total) * 100),
232
- });
233
- }
234
-
235
- const body = new Uint8Array(loaded);
236
- let position = 0;
237
- for (const chunk of chunks) {
238
- body.set(chunk, position);
239
- position += chunk.length;
240
- }
241
-
242
- return new Response(body, {
243
- headers: response.headers,
244
- status: response.status,
245
- statusText: response.statusText,
246
- });
247
- }
248
-
249
- /**
250
- * Clear cached models
251
- */
252
- async clearCache() {
253
- if (typeof caches !== 'undefined') {
254
- await caches.delete(this.cacheStorage);
255
- console.log('Model cache cleared');
256
- }
257
- }
258
-
259
- /**
260
- * List available models
261
- */
262
- static listModels() {
263
- return Object.entries(MODELS).map(([key, config]) => ({
264
- id: key,
265
- ...config,
266
- }));
267
- }
268
- }
269
-
270
- /**
271
- * Quick helper to create an embedder with a pre-configured model
272
- *
273
- * @example
274
- * ```javascript
275
- * import { createEmbedder } from './loader.js';
276
- *
277
- * const embedder = await createEmbedder('all-MiniLM-L6-v2');
278
- * const embedding = embedder.embedOne("Hello world");
279
- * ```
280
- */
281
- export async function createEmbedder(modelName = DEFAULT_MODEL, wasmModule = null) {
282
- // Import WASM module if not provided
283
- if (!wasmModule) {
284
- wasmModule = await import('./pkg/ruvector_onnx_embeddings_wasm.js');
285
- await wasmModule.default();
286
- }
287
-
288
- const loader = new ModelLoader();
289
- const { modelBytes, tokenizerJson, config } = await loader.loadModel(modelName);
290
-
291
- const embedderConfig = new wasmModule.WasmEmbedderConfig()
292
- .setMaxLength(config.maxLength)
293
- .setNormalize(true)
294
- .setPooling(0); // Mean pooling
295
-
296
- const embedder = wasmModule.WasmEmbedder.withConfig(
297
- modelBytes,
298
- tokenizerJson,
299
- embedderConfig
300
- );
301
-
302
- return embedder;
303
- }
304
-
305
- /**
306
- * Quick helper for one-off embedding (loads model, embeds, returns)
307
- *
308
- * @example
309
- * ```javascript
310
- * import { embed } from './loader.js';
311
- *
312
- * const embedding = await embed("Hello world");
313
- * const embeddings = await embed(["Hello", "World"]);
314
- * ```
315
- */
316
- export async function embed(text, modelName = DEFAULT_MODEL) {
317
- const embedder = await createEmbedder(modelName);
318
-
319
- if (Array.isArray(text)) {
320
- return embedder.embedBatch(text);
321
- }
322
- return embedder.embedOne(text);
323
- }
324
-
325
- /**
326
- * Quick helper for similarity comparison
327
- *
328
- * @example
329
- * ```javascript
330
- * import { similarity } from './loader.js';
331
- *
332
- * const score = await similarity("I love dogs", "I adore puppies");
333
- * console.log(score); // ~0.85
334
- * ```
335
- */
336
- export async function similarity(text1, text2, modelName = DEFAULT_MODEL) {
337
- const embedder = await createEmbedder(modelName);
338
- return embedder.similarity(text1, text2);
339
- }
340
-
341
- export default {
342
- MODELS,
343
- DEFAULT_MODEL,
344
- ModelLoader,
345
- createEmbedder,
346
- embed,
347
- similarity,
348
- };
@@ -1,21 +0,0 @@
1
- MIT License
2
-
3
- Copyright (c) 2025 rUv
4
-
5
- Permission is hereby granted, free of charge, to any person obtaining a copy
6
- of this software and associated documentation files (the "Software"), to deal
7
- in the Software without restriction, including without limitation the rights
8
- to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
- copies of the Software, and to permit persons to whom the Software is
10
- furnished to do so, subject to the following conditions:
11
-
12
- The above copyright notice and this permission notice shall be included in all
13
- copies or substantial portions of the Software.
14
-
15
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
- IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
- FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
- AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
- LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
- OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
- SOFTWARE.