Assistant Configuration
Last modified: 08/03/2026
Overview
Sprig settings are grouped by mode so you can control privacy, prompts, and output behavior.
Core settings
| Setting | What it controls |
|---|---|
| Enable Sprig | Turns assistant features on/off globally |
| AI provider | Which backend receives requests |
| Model | Model name used for requests |
| Endpoint override | Custom provider base URL (custom provider only) |
| API key | Provider credential stored locally |
| Save chat history | Restores prior Ask/Review chats per note |
Ask mode settings
| Setting | What it controls |
|---|---|
| Custom instructions | Prompt rules for Ask responses |
| Include images from note in messages | Sends embedded images with Ask requests |
Review mode settings
| Setting | What it controls |
|---|---|
| Custom instructions | Prompt rules for Review responses |
| Include images from note in messages | Sends embedded images with Review requests |
Generate mode settings
| Setting | What it controls |
|---|---|
| Custom instructions | Prompt rules for generated cards |
| Include images from note in messages | Sends embedded images for generation (vision-capable models required) |
| Approximate number of cards | Target count (1-10) |
| What flashcard types to generate | Enable/disable Basic, Reversed, Cloze, MCQ, Ordered, IO |
| Generated fields | Include/exclude T, I, and G output rows |
Model image capability notes
- Some models cannot analyse image content.
- More advanced models can identify text in images and convert it into question cards.
- The most advanced models may attempt to generate image occlusion (IO) cards.
- IO mask positioning can still be imperfect, so open generated IO cards in the flashcard editor and verify mask placement before studying.
Suggested starting profile
- Keep card count low (3-5) while tuning prompts.
- Start with text card types (Basic/Cloze) before enabling all types.
- Keep image input disabled unless you are intentionally using a vision model.
- Keep custom instructions concise and test in one note first.
Safety and policy
All generated content should be reviewed before use. See AI Usage Policy.