I like the concept, but I think a more reliable/less compute intensive way to implement it would be too use AI to call up non -AI data. I could just type in "some red beans and rice" and the LLM parses what I mean, and retrieves stored verified data.
That's what OpenNutrition does. However, in many cases, there is no publicly accessible "non-AI data" source to refer to. OpenNutrition tries to bridge the gap, using the public data when available, and providing additional inferred data to fill in the gaps. For "red beans" and "rice", OpenNutrition provides a long list of foods with full citations in public databases. See the "References" section where you can click through to the source material.
Red Beans
- https://www.opennutrition.app/search/red-beans-canned-and-dr...
- https://www.opennutrition.app/search/red-beans-dry-vIh9Ofhcl...
Rice
- https://www.opennutrition.app/search/enriched-white-rice-tlA...