Mistral Vision OCR for Knowledge Base

Turn Images & Handwriting
Into Searchable Knowledge

Use Mistral AI's vision models to extract text from images, receipts, handwritten notes, and scanned documents. Make everything searchable in your knowledge base.

Vision AI
OCR Parsing
Auto-Extract

The Image Problem

Your knowledge base can search text documents, but what about images, scanned receipts, handwritten notes, and photos of documents? They're invisible to search.

Without OCR

  • Images are just binary blobs
  • Handwritten notes can't be searched
  • Scanned documents are dead weight
  • Receipts and invoices unusable
  • Knowledge stays locked in images

With Mistral OCR

  • AI extracts text from any image
  • Handwriting becomes searchable
  • Scanned docs fully indexed
  • Receipt data automatically parsed
  • Everything is findable

What Can It Parse?

Mistral's vision AI handles virtually any image with text

Handwritten Notes

Meeting notes, to-do lists, sticky notes, journal entries - any handwriting style

Scanned Documents

PDFs from scanners, faxes, photocopies, and document images with any layout

Receipts & Invoices

Extract data from receipts, invoices, bills, and financial documents

Screenshots

UI screenshots, error messages, dashboards, charts - extract all visible text

Product Labels

Packaging text, ingredient lists, warning labels, serial numbers

Forms & Tables

Structured data from forms, spreadsheets, tables, and grid layouts

How It Works

Automatic OCR processing powered by Mistral AI vision models

1

Upload Image

Add image to your knowledge collection (JPEG, PNG, PDF, etc.)

2

Select Mistral OCR

Choose Mistral OCR parser for the resource

3

AI Extracts Text

Mistral vision model reads and converts to markdown

4

Now Searchable

Text is chunked, embedded, and ready for AI search

See It In Action

Actual handwritten grocery list parsed by Mistral OCR

Original Image

Handwritten grocery list

Extracted Text

- potatoes
- peas & carrots
- pastina
- garbage bags
- dog treats
- aluminum foil
- almond milk
- creamer - vanilla
- eggs (2)
- crushed tomatoes
- hot sauce
- paper towels?

Result: The handwritten list is now fully searchable in your knowledge base. Ask your AI assistant "What items are on the grocery list?" and it will find this document and list all items.

How to Set It Up

Configure Mistral OCR for your knowledge base in minutes

1

Set Up Mistral OCR Models

The module comes pre-configured with Mistral's OCR models. View them under LLM → Configuration → Models, filtered by "ocr".

Mistral OCR Models

Available OCR Models:

  • mistral-ocr-latest: Latest OCR model (recommended)
  • mistral-ocr-2505: May 2025 version
  • mistral-ocr-2503: March 2025 version
Note: Set up the Mistral provider via llm_mistral module and click "Fetch Models" to download the available OCR models from Mistral AI.
2

Use Mistral OCR Parser

When creating or editing a knowledge resource, select "Mistral OCR Parser" and choose your preferred OCR model. Upload images and the parser will automatically extract text.

Mistral Parser Configuration

Parser Configuration:

  • Parser: Select "Mistral OCR Parser"
  • Provider: Mistral AI (auto-selected)
  • OCR Model: Choose mistral-ocr-latest or specific version
  • Supported formats: Images (PNG, JPG, WEBP), PDFs, scanned documents

Process & Start Searching!

Click "Process Resources" to extract text from your images. Once processed, all text becomes searchable through your AI assistant. Ask questions about the content and get instant answers with source citations.

Processing Pipeline

When you process an image resource:

  1. Mistral OCR Parser sends image to Mistral AI vision model
  2. Vision model analyzes image and extracts all text content
  3. Extracted text is saved to the resource's content field
  4. Text is chunked using your collection's chunker settings
  5. Chunks are embedded and stored in your vector database
  6. AI can now search and cite this content in responses

Real-World Use Cases

How teams use Mistral OCR to unlock knowledge from images

Expense Management

Upload receipt photos, extract vendor, amount, date, and items automatically for searchable expense records

"Find all Starbucks receipts from last month"

AI: Found 8 receipts totaling $127.50 (Sources: receipt_001.jpg through receipt_008.jpg)

Meeting Notes Archive

Scan handwritten meeting notes and make every decision, action item, and idea searchable

"What did we decide about the Q4 budget?"

AI: "Approved $50K increase for marketing" (Source: Meeting Notes Oct 15, 2024)

Legacy Document Digitization

Convert old scanned contracts, faxes, and archived paperwork into searchable digital knowledge

"Find the 2015 lease agreement terms"

AI: "5-year lease at $2500/month" (Source: Scanned Lease Agreement 2015)

Product Catalog

Extract product specs, ingredients, and details from packaging photos to build searchable catalogs

"Which products contain wheat?"

AI: Found 12 products with wheat in ingredients (Sources: Product labels from catalog)

Powerful Features

Everything you need for vision-based knowledge extraction

Multi-Language Support

Extract text in multiple languages including English, French, Spanish, and more

Markdown Output

Extracted text formatted as clean markdown for better chunking and retrieval

Table Extraction

Preserves table structure and relationships when extracting from forms and spreadsheets

Multi-Page Processing

Handles multi-page PDFs and image sets with proper page organization

Image Attachment Handling

Embedded images are saved as Odoo attachments with proper references

Seamless Integration

Works with existing llm_knowledge collections - just select the Mistral OCR parser

Quick Setup

Get started in 4 steps

1

Install Dependencies

Requires llm_knowledge and llm_mistral modules

2

Install This Module

Search for "LLM Knowledge Mistral" in Apps and click Install

3

Set Up Mistral Provider

Go to LLM → Configuration → Providers

  • Configure your Mistral AI provider with API key
  • Click "Fetch Models" to download available OCR models
4

Use Mistral OCR Parser

When adding images to collections, select "Mistral OCR Parser" and choose an OCR model