Skip to content
Sırdaş

PDF to Markdown for Mistral

Prepare your PDF for Mistral: clean Markdown, redacted personal data and a token count. All on your device.

Sent to servers: 0 BNothing left your device

About this tool

Mistral answers better with well-structured text: headings, lists and paragraphs, without repeated headers or stray page numbers. We turn your PDF into Markdown with that structure and estimate the tokens so you know whether it fits in one conversation.

Before you paste anything into Mistral, personal data is replaced with consistent placeholders ([NOMBRE_1], [DOCUMENTO_1]…): the same value always gets the same placeholder, so the model understands the document without knowing the people.

What Mistral says about the files you upload

Quoted from its own pages, without summarising what they do not say. Policies change: open the source before you decide.

WhatWhat it saysSource
Training on your dataLe Chat is now called Vibe
«Vibe: users are not opted out by default and can opt out at any time in their settings»
help.mistral.ai
How to turn it offpersonal accounts
«Under the Privacy section, disable the toggle labelled Allow your interactions to be used to train our models»
help.mistral.ai
Retentionpersonal accounts
«we may retain different types of data in our backend systems for different amounts of time»
help.mistral.ai

Checked on · Not found on its official pages: file limits, human review. We do not claim it.

When you do not need Sırdaş

If the document holds nobody's data and Mistral accepts the PDF, upload it directly: it is faster. Sırdaş is for PDFs with names, ID numbers or accounts, or when you do not want the whole file to leave your machine.

How it works

  1. 01

    Drop the PDF.

  2. 02

    Review the Markdown and the highlighted data.

  3. 03

    Copy (or download the .md) and paste it into Mistral.

Questions

Can't Mistral read the PDF directly?
Many assistants accept PDFs, but uploading sends the file to their servers. With Sırdaş the PDF never leaves your machine: you only share with Mistral the anonymized text you choose to paste.
Which data is hidden before pasting into Mistral?
ID and passport numbers, tax IDs, emails, phones, dates, addresses, cards, medical record numbers and labelled names (patient, landlord, signed by…). Detection is automatic and can miss items: review the result.
How do I know the token count?
We show an estimate (about 4 characters per token). It is indicative; each model counts slightly differently.

More tools