/**
 * Qdrant vector backend for iranti.
 *
 * Implements the `VectorBackend` interface against a Qdrant HTTP API server.
 * Suitable for deployments that need high-throughput vector search separate
 * from the primary Postgres database.
 *
 * Configuration (via `QdrantConfig` or env vars in the factory):
 *  - `url`        — Qdrant server base URL (e.g. http://localhost:6333)
 *  - `apiKey`     — optional API key header (`api-key`)
 *  - `collection` — collection name (default: iranti_facts)
 *
 * The collection is created lazily on the first upsert with `EMBEDDING_DIMENSIONS`
 * dimensions and cosine distance. Subsequent calls skip collection creation
 * via the `collectionReady` flag.
 */
import { VectorBackend, VectorConsistencyFilter, VectorMutationDbClient, VectorSearchResult, VectorUpsertParams } from '../vectorBackend';
type QdrantConfig = {
    url: string;
    apiKey?: string;
    collection: string;
};
export declare class QdrantBackend implements VectorBackend {
    private readonly baseUrl;
    private readonly apiKey?;
    private readonly collection;
    private collectionReady;
    constructor(config: QdrantConfig);
    private request;
    private ensureCollection;
    upsert(params: VectorUpsertParams, _db?: VectorMutationDbClient): Promise<void>;
    delete(id: string, _db?: VectorMutationDbClient): Promise<void>;
    search(vector: number[], topK: number, filter?: Record<string, unknown>): Promise<VectorSearchResult[]>;
    listIndexedIds(filter?: VectorConsistencyFilter, pageSize?: number): Promise<string[]>;
    ping(): Promise<boolean>;
}
export {};
//# sourceMappingURL=qdrantBackend.d.ts.map