The technology behind Belly Melter.
Belly Melter AI achieves its goal through the usage of a number of technologies.
The Web Application Firewall (WAF) is covered using Cloudflare.
The AI system is connected to the Mistral API and employs the new mistral-small-latest model. This model enhances the results compared to the older GPT-4 version and is used to generate dietary information based on individual measurements, specifically focusing on Components and the AI Diet Meal Plan.
The payload supplied to the Mistral model is always the same; the only thing that varies is the user input data, which is updated each time the user fills out the Getting Started form.
This website does not save any user information, and no cookies have been harmed in the making of this "movie" :). During the connection between the client (user) and the assistant, only session (PHPSESSID) information is used.
The backend is exclusively developed using PHP, while the frontend utilizes Bootstrap framework and makes use of AJAX requests for dynamic functionality.
The submission form is controlled by Google reCaptcha V3.
The PDF file is generated using the fpdf.org class.
The logo has been generated using MidjourneyAI.
I have improved the Spanish translation capabilities of bellymelter by using the mistral-small-latest model to directly translate the output into Spanish. Previously, I had been relying on the Google Translator API for this task. However, I have found that using the mistral-small-latest model for translation not only provides better quality translations but also allows for a more seamless integration with the rest of the project.
The code has been analized by Bright scanner from https://app.neuralegion.com. No security issues has be found between the code.