Hi Dolibarr community! I wanted to share an exciting update about the “AI Assistants” module that I’ve published on Dolistore this summer. The latest version now includes a feature that many of you might find interesting: the ability to generate images using AI directly within your conversations.
The module currently supports these amazing open-source models:
You as administrator can choose which models must be available (usable) for your dolibarr users. Because as you know some of them are note so cheap 
These models are accessible through fal.ai’s API, which exclusively offers open-source generative models. What’s really cool about this is the cost-effectiveness: using StableDiffusion Turbo, you can generate around 14,000 images for just $1 USD! Even with the higher-quality FLUX-SCHNELL model, you still get about 300 images per $1 USD.
Here’s how it works: during any text conversation with an AI assistant, you can temporarily switch to an image generation model and request an image. The generation is instant, and the image is stored right there in your conversation thread. So you will have stored those images in your Dolibarr server although you don’t user any more the services of fal.ai ! Even better: in a same place you have the images generated through different models/providers.
Quick tips:
- To keep your server space optimized, consider deleting any images you don’t need using the “trash” icon (images typically range from 500KB to 1.1MB)
- Pro tip: for best results, try writing your prompts in English!
- To include text in the image: FLUX is almost perfect doing it with english text, but with other languages it runs better avoiding this:
- long sentences
- brand names or personal names not existing in the dictionary
but sometimes it works fine:
Feel free to share your experiences or ask any questions. I’m always here to help! 
