types.ts

export interface EmbeddingProvider {
  readonly name: string;
  readonly modelName: string;
  readonly dimensions: number;

  initialize(): Promise<void>;
  embed(text: string): Promise<number[]>;
  embedBatch(texts: string[]): Promise<number[][]>;
  isReady(): boolean;
}

export interface EmbeddingConfig {
  provider: 'transformers' | 'ollama' | 'openai' | 'custom';
  model?: string;
  batchSize?: number;
  maxRetries?: number;
  apiKey?: string;
  apiEndpoint?: string;
}

export function parseEmbeddingProviderName(
  value: unknown
): EmbeddingConfig['provider'] | undefined {
  if (value === 'transformers' || value === 'ollama' || value === 'openai' || value === 'custom') {
    return value;
  }
  return undefined;
}

// Default: bge-small (fast, ~2min indexing, consumer-hardware safe)
// Opt-in: set EMBEDDING_MODEL=onnx-community/granite-embedding-small-english-r2-ONNX for
// better conceptual search at the cost of 5-10x slower indexing and higher RAM usage
export const DEFAULT_MODEL = process.env.EMBEDDING_MODEL || 'Xenova/bge-small-en-v1.5';

function getDefaultApiEndpoint(provider: EmbeddingConfig['provider']): string | undefined {
  if (provider === 'ollama') {
    return process.env.OLLAMA_HOST || 'http://localhost:11434';
  }
  if (provider === 'openai') {
    return process.env.OPENAI_BASE_URL || 'https://api.openai.com/v1';
  }
  return undefined;
}

export const DEFAULT_EMBEDDING_CONFIG: EmbeddingConfig = {
  provider: parseEmbeddingProviderName(process.env.EMBEDDING_PROVIDER) ?? 'transformers',
  model: DEFAULT_MODEL,
  batchSize: 32,
  maxRetries: 3,
  apiKey: process.env.OPENAI_API_KEY,
  get apiEndpoint() {
    return getDefaultApiEndpoint(this.provider);
  }
};