RECENTLY CHECKED
All projects →
CAIBI suggestion · Physical AI

Doorbell that says who is there

Describe a visitor in words on a speaker in the house, such as a courier with a parcel, from a camera snapshot and a vision model.

Not yet assessed Not built by CAIBI

A source to explore and a proposed direction to develop. This is not a full build assessment.

Share your build

Make it useful to you

Describe a visitor in words on a speaker in the house, such as a courier with a parcel, from a camera snapshot and a vision model.

CAIBI’s proposed addition: Start with a local model if images of visitors should not leave the house; a cloud provider receives every snapshot. Keep an ordinary chime as the fallback.

Start with the original

Existing project

CAIBI suggestion developed from a source-reviewed catalogue project on 24 Sep 2026; untested

See the original source for its photos and demonstrations. We have not copied imagery without confirmed reuse rights. A source prototype is not a photo of this proposed CAIBI build.

Prompts to move it forward

Copy a prompt into your AI tool with the source material and your actual parts. Work through one small step at a time and check the output before using it. Choose an AI tool or practise the skills first.

Check the source before buying
Review the source material I supply for Doorbell that says who is there. Make a table of what is provided, missing, contradictory or dependent on a paid service. Identify exact hardware revisions, firmware files, library versions, CAD files and licences. Separate creator claims from your proposed additions. Do not fill gaps with guessed specifications. Finish with one small test that can resolve the biggest uncertainty before I buy parts.

Sources:
github.com: https://github.com/valentinfrlch/ha-llmvision
Plan the next step
I want to develop this idea: Doorbell that says who is there. Intended outcome: Describe a visitor in words on a speaker in the house, such as a courier with a parcel, from a camera snapshot and a vision model. The source is https://github.com/valentinfrlch/ha-llmvision. CAIBI suggests: Start with a local model if images of visitors should not leave the house; a cloud provider receives every snapshot. Keep an ordinary chime as the fallback. First inspect the source I supply. Separate what it already does from proposed additions. Ask for my actual board, parts, skill level and missing files. Produce a small staged plan with one bench test per stage. Do not invent pin assignments, dimensions or completed test results.

Sources:
github.com: https://github.com/valentinfrlch/ha-llmvision
Design a measured enclosure
Help me design a practical enclosure for Doorbell that says who is there. First ask for measured component dimensions, mounting holes, connectors, cable clearance, ventilation and operating conditions. Suggest a printable or off-the-shelf approach. Begin with a fit-test piece before a complete case. Label assumptions and do not invent an enclosure rating.

Sources:
github.com: https://github.com/valentinfrlch/ha-llmvision
Check and document the result
Create a test log for Doorbell that says who is there. Intended outcome: Describe a visitor in words on a speaker in the house, such as a courier with a parcel, from a camera snapshot and a vision model. Separate expected behaviour from observed results. Include power restart, disconnected inputs, communication failure and the actual physical use. Ask me to provide results, firmware versions, changes, photos and unresolved problems. Never mark a test passed without my evidence.

Sources:
github.com: https://github.com/valentinfrlch/ha-llmvision
Community ideas & builds

What builders are contributing

Suggest something useful, improve a route or share what you made. Contributions are reviewed before publication. Community builds remain attributed to their builders.

Contribute to this idea

Loading reviewed contributions…