Food Search (AI-assisted) in Loop: A Highly Experimental Add-On
The open-source Loop community continues to innovate, with one of the newest experimental features being Food Search, assisted by AI. This can serve as a food-entry helper: allowing users to search for foods, scan barcodes, and even scan a plate to help estimate carbohydrate content.
As many of you know, I’ve been running Trio for a long time. Recently, I switched back to Loop to try this out!
What is Food Search assisted by AI?
Food Search (AI-assisted) is designed to streamline food logging in Loop:
- Food lookup with AI-assisted search.
- Barcode scanning to pull up foods quickly.
- Plate scanning (photo-based recognition) to suggest carbohydrate amounts.
It’s important to stress that this feature is not part of official Loop and does not automatically change Loop’s algorithm or dosing. All carb entries still need to be reviewed and confirmed by the user.
Development Status
This feature originated from a community pull request (PR #2329) and is still considered highly experimental.
It is not included in the standard Loop customization script. To enable it, users must follow manual instructions (via Mac-Xcode or browser builds). Development discussions, troubleshooting, and updates are taking place in the Loop Zulip chat under #development ? New Feature: Food Search 2.0 for Loop! #2329.
Plate-Scanning Options and Cost
Currently, there are three plate-scanning options, and all of them come with a small associated cost. Because each scan draws from an AI service, it’s important to be mindful of usage.
For example, when I set up Food AI in Loop I chose OpenAI (ChatGPT API). In set up, I added a $20 spending limit to ensure I didn’t go over budget while experimenting with the plate-scanning feature. This is a practical safeguard for anyone curious to try it without risking surprise charges. Also worth noting, If you have a paid chat gpt account, this is something separate from that.
Safety and Practical Considerations
- Highly experimental: If trying this, you should be an advanced loop user.
- Verify carb estimates: Always cross-check results with your own carb knowledge before dosing.
- Test cautiously: Try it first in safe, controlled settings (at home with familiar meals).
- Follow development updates: Stay engaged with the Zulip thread for fixes, feedback, and evolving guidance.
Food Search (AI-assisted) highlights the creativity of the Loop community and the potential for AI tools to make everyday diabetes management more convenient.
Still, it should be viewed strictly as an experimental helper tool and not a substitute for carb counting, not a clinically validated system, and not a guarantee of accuracy. Used thoughtfully, with spending limits and safety checks, it can be an interesting way to explore the future of food logging within Loop.
← Back to September 2025 Newsletter
