@croma-ai/parse

Parse documents effortlessly.

Installation

pnpm add @croma-ai/parse

Quick Start

Set CROMA_API_KEY in your environment, then:

import { croma, parse } from "@croma-ai/parse";

const result = await parse({
  model: croma("mistral-ocr"),
  url: "https://example.com/document.pdf",
});

result.outputs?.markdown;  // extracted text
result.stats;              // { time, cost, tokens, pages }

Parse from File

import { croma, parse } from "@croma-ai/parse";

const file = new File([buffer], "document.pdf", { type: "application/pdf" });

const result = await parse({
  model: croma("mistral-ocr"),
  file,
});

result.outputs?.markdown;          // extracted text
result.outputs?.json?.blocks;      // structured blocks with bboxes
result.stats;                      // { time, cost, tokens }

Available Models

| Model ID | Provider | Category | | ----------------- | ----------- | --------------- | | mistral-ocr | Mistral | Document Parser | | llamaparse | LlamaIndex | Document Parser | | datalab-marker | Datalab | Document Parser | | gpt-4o | OpenAI | Vision LLM | | claude-sonnet-4 | Anthropic | Vision LLM | | gemini-2-flash | Google | Vision LLM |

Configuration

const croma = createParse({
  apiKey: process.env.CROMA_API_KEY,   // or set CROMA_API_KEY env var
  baseURL: "https://parse.croma.run",   // default
});

Types

All types are exported:

import type {
  ParseResult,
  ParseOutputs,
  ParseStats,
  ParseBlock,
  BBox,
  ParseModel,
  ParseProvider,
} from "@croma-ai/parse";

License

MIT