docuglean-ocr/dist/types.d.ts

An SDK for intelligent document processing using State of the Art AI models.

import { z } from 'zod';
export type Provider = 'openai' | 'mistral' | 'gemini';
export declare const validateConfig: (config: OCRConfig | ExtractConfig) => void;
interface MistralOCRImage {
    id: string;
    topLeftX: number | null;
    topLeftY: number | null;
    bottomRightX: number | null;
    bottomRightY: number | null;
    imageBase64?: string | null;
}
interface MistralOCRPage {
    index: number;
    markdown: string;
    images: MistralOCRImage[];
    dimensions: {
        dpi: number;
        height: number;
        width: number;
    } | null;
}
export interface MistralOCRResponse {
    pages: MistralOCRPage[];
}
export interface OpenAIOCRResponse {
    text: string;
    usage: {
        prompt_tokens: number;
        completion_tokens: number;
        total_tokens: number;
    };
}
export interface GeminiOCRResponse {
    text: string;
    model_used: string;
}
export interface OCRPage {
    index: number;
    markdown: string;
}
export interface OCRConfig {
    filePath: string;
    provider?: Provider;
    model?: string;
    apiKey: string;
    options?: {
        mistral?: {
            includeImageBase64?: boolean;
        };
        openai?: {
            maxTokens?: number;
        };
        gemini?: {
            temperature?: number;
            topP?: number;
            topK?: number;
        };
    };
    prompt?: string;
}
export interface ExtractConfig {
    filePath: string;
    apiKey: string;
    provider?: Provider;
    model?: string;
    prompt?: string;
    responseFormat?: z.ZodType<any>;
    systemPrompt?: string;
}
export interface BaseStructuredOutput {
    [key: string]: any;
}
export interface StructuredExtractionResult<T extends BaseStructuredOutput> {
    raw: string;
    parsed: T;
}
export interface OCRResult {
    markdown: string;
    images: any[];
    rawResponse: MistralOCRResponse;
}
export {};