Embeddings
Configure embedding models for vector-based memory search.
createEmbeddingModel
import { createEmbeddingModel } from '@ahzan-agentforge/core';
const embedding = createEmbeddingModel({
provider: 'openai',
model: 'text-embedding-3-small',
dimensions: 1536,
});EmbeddingConfig
interface EmbeddingConfig {
provider?: string; // 'openai' or other providers
model?: string; // Model identifier
dimensions?: number; // Vector dimensions
}EmbeddingModel Interface
interface EmbeddingModel {
embed(text: string): Promise<number[]>;
embedBatch(texts: string[]): Promise<number[][]>;
dimensions: number;
}Usage with PgVector
The embedding model is used automatically by PgVectorMemoryStore for both storing and searching:
const store = await createPgVectorStore({
connectionString: process.env.DATABASE_URL,
embedding: {
provider: 'openai',
model: 'text-embedding-3-small',
},
});