/**
 * Default configuration values
 *
 * Environment isolation:
 * - Default Vectorize index is the DEV index for safety
 * - Production indexing must explicitly specify --index classic-chat-knowledge
 * - Override via VECTORIZE_INDEX_NAME environment variable or --index CLI flag
 *
 * Embedding Model: BGE-M3
 * - 1024 dimensions (richer semantic representation than bge-base's 768)
 * - 8192+ token context window (vs 512 for bge-base - no truncation risk)
 * - Multi-lingual support (100+ languages)
 * - $0.012/M tokens (5.5x cheaper than bge-base at $0.067/M)
 * - Supports dense, sparse, and ColBERT retrieval modes
 *
 * @see https://developers.cloudflare.com/workers-ai/models/bge-m3/
 * @see https://huggingface.co/BAAI/bge-m3
 */
/**
 * Default category patterns for common documentation and marketing structures.
 * Used by content sources to detect categories from paths.
 *
 * Patterns are matched against the path using includes(), so they work
 * with both URL paths (/warranty/standards/) and file paths (04-standards/).
 * Patterns without leading slashes are more flexible for matching.
 */
export declare const DEFAULT_CATEGORY_PATTERNS: Record<string, string>;
export declare const DEFAULTS: {
    chunking: {
        strategy: "heading";
        headingLevel: number;
        /** Minimum chunk size - increased from 50 for better quality chunks */
        minSize: number;
        /** Maximum chunk size in characters (~800-1000 tokens, well within bge-m3's 8192 limit) */
        maxSize: number;
    };
    embedding: {
        /**
         * BGE-M3: Multi-lingual, multi-functionality embedding model
         * Replaces bge-base-en-v1.5 for better quality and lower cost
         */
        model: string;
        /** Output dimensions - must match Vectorize index configuration */
        dimensions: number;
        /**
         * Batch size for API requests
         * Reduced from 50 to 20 to handle large content chunks better.
         * Large web pages can have 10-30KB chunks which timeout at batch size 50.
         */
        batchSize: number;
        /** Maximum tokens per text (bge-m3 supports 8192+) */
        maxTokens: number;
    };
    vectorize: {
        indexName: string;
        batchSize: number;
        autoInit: {
            enabled: boolean;
            dimensions: number;
            metric: "cosine";
            metadataIndexes: {
                propertyName: string;
                type: "string";
            }[];
        };
    };
    url: {
        pattern: string;
    };
};
//# sourceMappingURL=defaults.d.ts.map