import { createEmbeddingSemanticRanker, type EmbeddingProvider, type ProjectRepository, type SemanticRanker, } from '@portfolio/chat-data'; import { createOpenAIEmbeddingProvider } from './embedding';

const DEFAULT_PROJECT_EMBEDDING_MODEL = 'text-embedding-3-large'; const DEFAULT_PROJECT_SEMANTIC_SCORE = 12;

type CreateSemanticRankerOptions = { projectRepository: ProjectRepository; embeddingProvider?: EmbeddingProvider | null; getEmbeddingClient?: () => Promise<import('openai').OpenAI | null>; scoreScale?: number; embeddingModel?: string; };

export function createSemanticRanker(options: CreateSemanticRankerOptions): SemanticRanker { const { projectRepository, embeddingProvider, getEmbeddingClient, scoreScale, embeddingModel } = options;

const resolvedEmbeddingProvider = embeddingProvider ?? (getEmbeddingClient ? createOpenAIEmbeddingProvider({ model: embeddingModel?.trim() || DEFAULT_PROJECT_EMBEDDING_MODEL, getClient: getEmbeddingClient, logScope: 'chat-project-search', }) : { async embedTexts(texts: string[]): Promise<number[][]> { return texts.map(() => []); }, });

return createEmbeddingSemanticRanker({ embeddingProvider: resolvedEmbeddingProvider, projectRepository, scoreScale: typeof scoreScale === 'number' ? scoreScale : DEFAULT_PROJECT_SEMANTIC_SCORE, }); }

export type { SemanticRanker };