feat(core): give Waggle its own database, apart from memory (P1.5)
Waggle kept its own tables in the memory database. Before the memory
engine can be replaced by mem0 they need a home of their own.
- add WaggleDB (<dataDir>/waggle.db). Automations, notifications, held
actions, skill hashes, thumbs feedback, cost tracking, the profile card
and awareness items live there. Their stores take a WaggleDB, so passing
a memory database is a type error.
- the profile card and awareness items are now per scope ('personal' or
'workspace:<id>') in one table each, instead of one table per .mind file
- on the first start, copy these tables from personal.mind and every
workspace.mind (importLegacyMindData). The legacy files are only read,
each scope is copied once and a table only while its target is empty.
- copy the logger, injection scanner, external-memory ingress guard,
WorkspaceManager and the embedders from @waggle/hive-mind-core into
@waggle/core, which now re-exports its own copies
- point the server, agent orchestrator and CLI at waggle.db; update the
tests that seeded these tables in a memory database
ADR-0007 records the layout and the licensing question about the copied
files.
This commit is contained in:
73
packages/core/src/embeddings/inprocess-embedder.ts
Normal file
73
packages/core/src/embeddings/inprocess-embedder.ts
Normal file
@@ -0,0 +1,73 @@
|
||||
/**
|
||||
* In-process embedder using @huggingface/transformers (ONNX Runtime).
|
||||
* Default provider for ALL desktop users — zero config, works offline.
|
||||
* Model: Xenova/all-MiniLM-L6-v2 (384 native dims, normalized to target dims).
|
||||
* Downloads ~90MB fp32 model on first use, cached in ~/.waggle/models/.
|
||||
*/
|
||||
|
||||
import path from 'node:path';
|
||||
import os from 'node:os';
|
||||
import type { Embedder } from './embeddings.js';
|
||||
import { createCoreLogger } from '../logger.js';
|
||||
import { withTransformersModelLoad } from './transformers-model-load.js';
|
||||
|
||||
const log = createCoreLogger('inprocess-embedder');
|
||||
|
||||
export interface InProcessEmbedderConfig {
|
||||
model?: string;
|
||||
cacheDir?: string;
|
||||
targetDimensions?: number;
|
||||
}
|
||||
|
||||
/** Normalize embedding dimensions: zero-pad shorter, truncate longer. */
|
||||
export function normalizeDimensions(embedding: Float32Array, targetDims: number): Float32Array {
|
||||
if (embedding.length === targetDims) return embedding;
|
||||
const result = new Float32Array(targetDims);
|
||||
const copyLen = Math.min(embedding.length, targetDims);
|
||||
result.set(embedding.subarray(0, copyLen));
|
||||
return result;
|
||||
}
|
||||
|
||||
export async function createInProcessEmbedder(config?: Partial<InProcessEmbedderConfig>): Promise<Embedder> {
|
||||
const model = config?.model ?? 'Xenova/all-MiniLM-L6-v2';
|
||||
const cacheDir = config?.cacheDir ?? path.join(os.homedir(), '.waggle', 'models');
|
||||
const targetDims = config?.targetDimensions ?? 1024;
|
||||
|
||||
log.info(`Loading in-process embedding model: ${model} (~90MB fp32 first download)`);
|
||||
|
||||
const { pipeline } = await import('@huggingface/transformers');
|
||||
const extractor = await withTransformersModelLoad({
|
||||
cacheDir,
|
||||
model,
|
||||
load: (canonicalCacheDir) => pipeline('feature-extraction', model, {
|
||||
dtype: 'fp32',
|
||||
cache_dir: canonicalCacheDir,
|
||||
}),
|
||||
onQuarantine: () => log.warn(`Quarantined corrupt embedding model cache: ${model}`),
|
||||
});
|
||||
const nativeDims = 384; // all-MiniLM-L6-v2 output dimensions
|
||||
|
||||
log.info(`In-process embedder ready (${nativeDims} native dims → ${targetDims} normalized)`);
|
||||
|
||||
return {
|
||||
dimensions: targetDims,
|
||||
|
||||
async embed(text: string): Promise<Float32Array> {
|
||||
const result = await extractor(text, { pooling: 'mean', normalize: true });
|
||||
const raw = new Float32Array(result.data as Float32Array);
|
||||
return normalizeDimensions(raw, targetDims);
|
||||
},
|
||||
|
||||
async embedBatch(texts: string[]): Promise<Float32Array[]> {
|
||||
if (texts.length === 0) return [];
|
||||
const results: Float32Array[] = [];
|
||||
// Process one at a time to avoid memory issues with large batches
|
||||
for (const text of texts) {
|
||||
const result = await extractor(text, { pooling: 'mean', normalize: true });
|
||||
const raw = new Float32Array(result.data as Float32Array);
|
||||
results.push(normalizeDimensions(raw, targetDims));
|
||||
}
|
||||
return results;
|
||||
},
|
||||
};
|
||||
}
|
||||
Reference in New Issue
Block a user