Optimizing your productivity has never been smoother! Thanks to the native integration of Mistral, the French open-source AI, directly within TimeTonic, you can automate complex tasks, extract data, analyze PDF documents or answer business-specific questions without leaving your workspace.
Why choose Mistral with TimeTonic?
✅ French Sovereignty: hosted in Europe, RGPD compliant...
✅ Performance: fast response, fine contextual understanding, JSON format support
✅ Confidentiality: your data never leaves your secure environment
✅ No-code simplicity: everything can be configured in just a few clicks without writing a single line of code
Start here ⬇️
1️⃣ Click the menu → ⚙️ Automation
To access the list of available automations.🟠
2️⃣ Select → Automation with Timetonic + AI
A new window opens, showing existing automations and allowing you to create a new one.
3️⃣ Click the button → ＋ Create an automation
A new window opens:
⚙️ This is the automation configuration form
1️⃣ Name the scenario
So you can easily identify it later
For example, "AI Response for client fields".
2️⃣ Select a trigger to launch the automation.
When: the field is modified.
3️⃣ Add an additional condition if needed
4️⃣ Click the condition → "Then"
To access the automation actions window
5️⃣ Select "Ask Mistral AI"
A new window opens:
⚙️ The form to connect your AI and configure its role and actions.
1️⃣ API Key
Insert your personal API key to allow TimeTonic to communicate with Mistral.
👉 If you don't have one yet, go here to create or retrieve it: Mistral API Keys
❗A valid API key is required. Requires an active Mistral plan.
2️⃣ Organization ID (optional)
Enter your Mistral organization ID if you are using one (e.g., in a corporate or shared team environment).
3️⃣ Model
Select the Mistral AI model depending on the task to automate.
Here’s a quick overview:
| Model | Recommended Use |
|---|---|
mistral-tiny |
⚡ Fast & lightweight – Perfect for classification, short generation, or simple analysis |
mistral-small |
🧠 Moderate reasoning – Ideal for data extraction, content summarization, or drafting |
mistral-medium |
🔍 More powerful – For complex or in-depth analysis needs |
ℹ️ Check the Mistral documentation to see the latest available versions.
4️⃣ AI Role Description
Describe the role the AI should play (also called “system prompt”).
"You are an OCR specialist for bank documents like RIB (bank account identity)."
5️⃣ Question
This is where you provide your direct instruction to the AI. You can include a dynamic field using the character $.
Example:
"You must extract from this image the banking information in JSON format in the following fields:
IBAN: Format the IBAN with spaces every 4 characters for readability
BIC: The BIC / Swift code
CodeBanque: Bank code
CodeGuichet: Branch code
NumCompte: Bank account number
CLE: Account key
Domiciliation: Text form of the domicile
Titulaire: Account holder name
AdresseTitulaire: Account holder address in text
Devise: Currency"
6️⃣ Attachment (optional)
If you're analyzing an image or a PDF, indicate the file field containing the attachment.
This is what Mistral will use as input.
Example: RIB image
7️⃣ Maximum tokens
Defines the maximum number of tokens the AI can use to reply (e.g., 500 or 1000).
⚠️ The higher the number, the longer (and potentially more costly) the response will be.
8️⃣ Choose the response format
- Text: Stores the full response as plain text in one field (SmartText or a multiline text field recommended) for basic answers.
- JSON with field mapping: The response is structured in JSON, each property is automatically distributed to the appropriate fields. (Used for data extraction from images)
9️⃣ Define the names to extract in JSON from your image in the "JSON Property Name" fields (should match what's in your prompt)
🔟 Choose the fields
Select the primary destination field in your table to store the AI's response.
If you chose JSON: make sure to create the destination fields for each property beforehand (same name and syntax as your expected JSON).
⚠️ Tip: Naming consistency is key
💡 For the JSON response to be correctly distributed across fields, the names must strictly match:
- In the prompt (field 5️⃣)
- In the JSON keys
- In the TimeTonic column names
Avoid spaces, accents, inconsistent capitalizations.
See the article: