/** * CLI embedder — pluggable text-to-vector embedding for search commands * @generated by @constructive-io/graphql-codegen * DO NOT EDIT - changes will be overwritten */ // ─── Types ─────────────────────────────────────────────────────────────────── export type EmbedderFunction = (text: string) => Promise; export interface EmbedderConfig { /** Provider name: 'ollama' or 'custom' */ provider: string; /** Model identifier (e.g. 'nomic-embed-text') */ model?: string; /** Base URL for the provider (e.g. 'http://localhost:11434' for Ollama) */ baseUrl?: string; } // ─── Built-in Ollama Provider ──────────────────────────────────────────────── function createOllamaEmbedder( baseUrl: string = 'http://localhost:11434', model: string = 'nomic-embed-text' ): EmbedderFunction { let clientP: | Promise<{ generateEmbedding: (text: string, model: string) => Promise<{ embedding: number[] }>; }> | undefined; return async (text: string): Promise => { if (!clientP) { clientP = import('@agentic-kit/ollama') .then((m) => new m.OllamaClient(baseUrl)) .catch(() => { throw new Error( 'The ollama embedder requires @agentic-kit/ollama. Install it: npm i @agentic-kit/ollama' ); }); } const client = await clientP; const result = await client.generateEmbedding(text, model); return result.embedding; }; } // ─── Embedder Resolution ───────────────────────────────────────────────────── /** * Resolve an embedder function from environment variables or appstash config. * * Resolution order: * 1. EMBEDDER_PROVIDER env var (+ EMBEDDER_MODEL, EMBEDDER_BASE_URL) * 2. appstash config keys: embedder.provider, embedder.model, embedder.baseUrl * 3. null (no embedder configured) * * @param store - Optional appstash config store for reading persisted config * @returns An EmbedderFunction or null if no embedder is configured */ export function resolveEmbedder(store?: { getVar: (key: string) => string | undefined; }): EmbedderFunction | null { // 1. Check environment variables first const envProvider = process.env.EMBEDDER_PROVIDER; if (envProvider) { return buildEmbedder({ provider: envProvider, model: process.env.EMBEDDER_MODEL, baseUrl: process.env.EMBEDDER_BASE_URL, }); } // 2. Check appstash config if (store) { const configProvider = store.getVar('embedder.provider'); if (configProvider) { return buildEmbedder({ provider: configProvider, model: store.getVar('embedder.model'), baseUrl: store.getVar('embedder.baseUrl'), }); } } // 3. No embedder configured return null; } /** * Build an embedder function from a config object. */ function buildEmbedder(config: EmbedderConfig): EmbedderFunction | null { switch (config.provider) { case 'ollama': return createOllamaEmbedder(config.baseUrl, config.model); default: console.error(`Unknown embedder provider: '${config.provider}'. Supported: ollama`); return null; } } /** * Auto-embed text values in vector where-clause fields. * * When --auto-embed is passed, any vector field in the where clause that * contains a text string (instead of a float array) will be converted to * an embedding vector using the configured embedder. * * @param where - The where clause object (mutated in place) * @param vectorFieldNames - Names of vector embedding fields (e.g. ['vectorEmbedding']) * @param embedder - The resolved embedder function * @returns The modified where clause */ export async function autoEmbedWhere( where: T, vectorFieldNames: string[], embedder: EmbedderFunction ): Promise { const rec = where as unknown as Record; for (const fieldName of vectorFieldNames) { const fieldValue = rec[fieldName]; if (fieldValue && typeof fieldValue === 'object') { const input = fieldValue as Record; // If 'vector' is a string, embed it if (typeof input.vector === 'string') { const text = input.vector; const embedding = await embedder(text); input.vector = embedding; } } else if (typeof fieldValue === 'string') { // Shorthand: --where.vectorEmbedding "text" with --auto-embed // becomes { vector: [embedded], metric: 'COSINE' } const embedding = await embedder(fieldValue); rec[fieldName] = { vector: embedding }; } } return where; } /** * Auto-embed text values in mutation input data (create/update). * * When --auto-embed is passed on create or update, any vector field in * the input data that contains a text string will be converted to an * embedding vector using the configured embedder. * * Usage: * csdk article create --input.embedding "Machine learning concepts" --auto-embed * csdk article update --id xxx --input.embedding "Updated description" --auto-embed * * This is a CLI-only convenience — in production, database triggers or * a job queue should handle embedding generation. * * @param data - The mutation input data object (mutated in place) * @param vectorFieldNames - Names of vector embedding fields (e.g. ['embedding']) * @param embedder - The resolved embedder function * @returns The modified data object with text values replaced by vectors */ export async function autoEmbedInput( data: T, vectorFieldNames: string[], embedder: EmbedderFunction ): Promise { const rec = data as unknown as Record; for (const fieldName of vectorFieldNames) { const fieldValue = rec[fieldName]; if (typeof fieldValue === 'string') { // Text string → embed to vector array const embedding = await embedder(fieldValue); rec[fieldName] = embedding; } // If it's already an array (pre-computed vector), leave it as-is } return data; }