Feature Request: AI Image Recognition for Quick Item Addition via Phone Camera

[2oby]

What is the problem you are trying to solve with this feature?

Currently, adding items to Home Box requires manual entry of all details. For users with large inventories or when adding multiple items, this can become time-consuming. An AI image recognition workflow would significantly reduce friction and make inventory management more accessible, especially for less tech-savvy family members.

What is the solution you are proposing?

Add AI-powered image recognition capability that allows users to quickly add new items to their Home Box inventory by simply taking a photo with their phone camera. The system would automatically extract item details from the image and pre-populate the item creation form.

What alternatives have you considered?

No response

Additional context

No response

Contributions

• I have searched through existing ideas in the discussions to check if this is a duplicate

[tankerkiller125]

In the latest version we added barcode scanning, you can now add items by their barcodes from the store.

AI based recognition is very likely very far off, and would likely be implemented as a separate project with an integration given it has very different levels of skills and requirements to work properly.

[LokiMetaSmith]

you should be able to do this with a decent MCP server.

my attempt to do so is here.
https://github.com/LokiMetaSmith/Homebox-mcp

[richjdsmith]

AI models have come a long way in the last year. The appeal of being able to take a picture of all the items in a storage location and have the AI create items within that location would sure be nice. Any chance this can be reconsidered with the rise of cheap & good flash models? @tankerkiller125

[tankerkiller125]

We most likely at some point will add MCP for agents (a PR already exists with the basics), but the chances of having a built in AI system is still pretty slim. We're strong believers in being agnostic with 3rd party things (in other words we would feel compelled to support every major implementation, not just one of OpenAI or Anthropic or Gemini). Additionally, it would significantly add to the maintenance burden given how rapidly the AI space moves.

Additionally, personally, I have spent so much time and effort into AI harnesses and other random stuff for AI projects at work that I am frankly burned out on integrating AI with things at this point, so any effort into it would have to come from someone else.

[richjdsmith]

Super fair, and I really appreciate the thoughtful response. I was poking around the codebase, and am trying to see if there was a way to implement this as more of a plugin. I too and tired of the fragility of AI (I've been using every model via openrouter and anything I build that has AI, I am ensuring it works across models). With that in mind, and keeping development overhead down, could you suggest an approach that would be able to add the AI to the app, but as a plugin without messing with the main repo? I'd be happy to see if I can get that working, but would appreciate a bit of guidance if you have any thoughts on how you'd want to see a plugin implemented. Perhaps how @LokiMetaSmith implemented the MCP?

Thanks for all your work on this app so far, it is incredible.