Solo founder from India. Every AI app I built needed the same thing: remember the images a user uploaded, and answer questions about them later.
I kept rebuilding the same stack: store → extract → embed → retrieve → isolate per user.
So I turned it into an API. Two calls:
await behavr.remember({ file: image })
await behavr.ask({ question: "What did the customer upload?" })
No vector DB or vision pipeline to run yourself. Semantic retrieval across everything a user has stored, per-user isolation, self-serve signup, npm SDK (npm install behavr-sdk).
Where it honestly stands: it's live, I have 0 paying customers, and I just started telling people about it. Free tier is 50 images/month, no card. Video support is on paid plans.
Try it without signing up: https://behavr.in/#/playground
Docs: https://behavr.in/#/docs
Example app (clone and run): https://github.com/confuse57/behavr-example-app
What I'd love to know:
- Would you integrate something like this, or build it in-house?
- Are $29/mo (500 images) and $99/mo (3,000 images) sensible?
- What's the first thing that would make you close the tab?
Top comments (1)
The two-call interface is appealing, but the isolation contract would decide whether I could integrate it. How is the user identity bound to remember/ask, and does deleting an upload also delete extracted text and embeddings? I would want a test where two users ask the same question and neither can retrieve the other’s image. For the 500-image plan, clarifying monthly new uploads versus total retained images would also make the ongoing cost easier to compare with building it in-house.