Block: Mistral Parser Block

The Mistral Parser block extracts text and structure from PDF documents using Mistral's OCR API. Point it at a publicly accessible PDF URL or upload a PDF directly, then receive the content in Markdown, plain text, or raw JSON. Reach for it when you need to turn PDFs into clean, machine-readable text for downstream blocks.

Overview

Property	Value
Type	`mistral_parse`
Category	tools
Color	`#000000`

When to Use

Convert a PDF from a public URL into formatted Markdown for an LLM or knowledge base.
Extract plain text from uploaded contracts, invoices, or reports for further processing.
Pull raw JSON OCR output when you need page-level structure and metadata.
Parse only specific pages of a large PDF instead of the whole document.
Feed scanned or image-based PDFs through OCR to recover their text content.

Configuration

Select Input Method

Dropdown choosing how the PDF is supplied: PDF Document URL (url) or Upload PDF Document (upload).

PDF Document URL

A direct, publicly accessible URL to a PDF (shown when input method is url). Cloud storage share links (Google Drive, Dropbox) are not supported; use a direct download URL.

Upload PDF

File upload for a PDF (shown when input method is upload), accepting application/pdf up to 50MB.

Output Format

Dropdown selecting the result format: Markdown (Formatted) (markdown), Plain Text (text), or JSON (Raw) (json).

Specific Pages

Optional comma-separated, zero-based page list (e.g. 0,1,2). Leave empty to parse all pages.

API Key

Required Mistral API key used to authenticate the OCR request.

Inputs & Outputs

Inputs: inputMethod (string), filePath (string), fileUpload (json), apiKey (string), resultType (string), pages (string)
Outputs: content (string), metadata (json)

Tools

Mistral PDF Parser (mistral_parser) — Calls the Mistral OCR API (https://api.mistral.ai/v1/ocr) using the mistral-ocr-latest model to extract text from a PDF supplied as a URL or an uploaded file. Returns the extracted content in the requested format (markdown, text, or JSON) along with a metadata object that includes jobId, fileName, fileType, pageCount, model, resultType, processedAt, and (when available) usageInfo containing pagesProcessed and docSizeBytes.

YAML Example

mistral_parse_1:
  type: mistral_parse
  name: "Mistral Parser"
  inputs:
    inputMethod: "url"
    filePath: "https://example.com/document.pdf"
    resultType: "markdown"
    pages: "0,1,2"
    apiKey: "{{MISTRAL_API_KEY}}"
  connections:
    outgoing:
      - target: next-block-id

Mistral Parser Block