/**
 * Markdown content chunking
 */
import type { Chunk, ChunkContext } from './types.js';
import { type MarkdownPreprocessorOptions } from './preprocessor.js';
/**
 * Section represents a split portion of markdown content
 */
export interface Section {
    heading: string | null;
    content: string;
}
/**
 * Extended chunk context with preprocessing options
 */
export interface ChunkContextWithPreprocessing extends ChunkContext {
    /** Markdown preprocessing options */
    preprocessing?: MarkdownPreprocessorOptions;
}
/**
 * Chunk a markdown file into indexable pieces
 */
export declare function chunkMarkdownFile(filePath: string, content: string, context: ChunkContextWithPreprocessing): Chunk[];
/**
 * Split markdown content by heading level.
 *
 * Splits content into sections based on headings of the specified level.
 * Content before the first heading is assigned a null heading.
 *
 * @param content - Markdown content to split
 * @param level - Heading level to split on (e.g., 2 for ##)
 * @returns Array of sections with heading and content
 */
export declare function splitByHeading(content: string, level: number): Section[];
/**
 * Default max characters per chunk before paragraph-splitting kicks in.
 *
 * BGE-M3 truncates long inputs at embed time, silently dropping context. A
 * heading-only or H1-only document collapses into one giant section, so we cap
 * section size here as a second axis of splitting beyond headings.
 */
export declare const MAX_SECTION_CHARS = 4000;
/**
 * Split an oversized section into multiple sections along paragraph (then
 * sentence) boundaries. Sections at or under `maxSize` pass through untouched.
 * Continuation parts keep the heading, tagged "(continued)".
 */
export declare function splitOversizedSection(section: Section, maxSize?: number): Section[];
//# sourceMappingURL=markdown.d.ts.map