Back to all discussions

  •   about 1 month ago

Bee developer connect

Hi - is there somewhere I can ask a question about the Bee API / flow to their development / product team? Thanks, Andy

  • 16 comments

  •   •   about 1 month ago

    Hey Andy - Feel free to drop your questions about Bee here. I should be able to assist and can connect with the Bee team if needed.

    -Kourtney, Amazon

  •   •   about 1 month ago

    Thanks Kourtney - just one quick question actually. Is there a minimum duration for a recording, for it to be captured and processed? e.g. 10 seconds or anything like that? Or will any length of recording be captured and processed? Andy

  •   •   about 1 month ago

    That's a great question -- It needs to be greater than about 100 words or it won't be picked up

  •   •   about 1 month ago

    Hi Kourtney, thanks for reply. I don't think 100 words sounds right? I've recorded much less than that and it's picked it up. Thanks, Andy

  •   •   about 1 month ago

    Let me check in with the Bee engineers and I'll circle back

  •   •   about 1 month ago

    Hey Andy - The engineers said its now <50 words. Does that align with what you've seen?

  •   •   about 1 month ago

    Hey Kourtney — yes and no! When I record a short message via the Apple Watch app (I'm still waiting on my actual device to be delivered), it initially appears through the Bee API in the CAPTURING state. The real-time transcript is available, but end_time and the summary remain null. The recording sometimes never transitions to COMPLETED. Later, it disappears from the conversation feeds entirely, and requesting it directly by conversation ID returns 404 / Conversation not found. For example, a conversation contained “Add cucumber and relish to the grocery list.” I could retrieve it while it was CAPTURING, but it subsequently disappeared. Longer recordings seem to process and reach COMPLETED successfully, so I'm guessing there's threshold around the processing? It isn’t a blocker because I now snapshot the transcript into my own database when I first observe it. Something for the friction log!

  •   •   about 1 month ago

    This is currently expected behavior based on that need for it to be > 50 words. The idea behind is that anything shorter is often hallucinated from noise or far field speech so they are deleted. I can see how that's not ideal if you just want to give yourself a reminder like “Add cucumber and relish to the grocery list" so definitely put your feedback in your friction log so we can get it to the team for future improvement :)

  •   •   25 days ago

    Hi Kourtney - a Bee access request for a Bee-track entry.

    Project: Bystander (https://github.com/AtchayamG/bee-bystander). It does consent-aware redaction for people who are near a Bee wearer but never agreed to be recorded. The Bee integration is built on the documented bee proxy and the /v1 API.

    The blocker: I can't complete bee login.
    - I don't own a Bee Pioneer, and the Bee Android app won't go past device pairing without one.
    - Developer Mode is documented only for the iOS app, and I have only an Android phone (I'm in India).

    What I need (any ONE of these is enough):
    1. A developer token I can use with "bee login --token", or
    2. A sandbox or test Bee account the CLI can authenticate against, or
    3. A supported way to approve bee login without a Pioneer device.

    An empty account with no recordings is fine. I only need authenticated /v1/me and /v1/conversations calls to work, so judges can see real Bee API use and not only a 401. Any token stays out of the repo, logs and video.

    Timing: submissions close Oct 23, so anything in the next week or two would let me finish.

    Thanks,
    Atchayam G

  •   •   25 days ago

    Hey @atchayam - let me check in with the bee eng team and will follow up shortly

  •   •   22 days ago

    Hi Bee team — I’m participating in the hackathon and exploring a hosted integration with the Bee API. I tested the pairing protocol with my own account, but only using the CLI’s published app ID for a one-time local test. I don’t want to use that ID in an app.

    How can a third-party app obtain its own `app_id`? Does user approval require Developer Mode, or is there a supported flow for regular users? Also, where can I find guidance on token storage, expiration, refresh, revocation, and API usage limits?

    Thanks for pointing me to the right documentation or contact.

  •   •   11 days ago

    hi Kourtney Meiss, i'm participating in the amazon developer hackathon and i want to build for the bee track, but i don't own a bee pioneer and i'm an android user. the faq says the bee cli should not require a physical bee device, but it does require a bee account. i'm currently unable to authenticate without access to a pioneer. is there a supported way for hackathon participants without a bee device to get a developer/test account or authenticate the cli? i only need access to legitimate bee data for development and to demonstrate the bee cli/mcp integration. i don't need access to anyone else's data. thanks!

  •   •   11 days ago

    Bee track — Developer Mode/access without Bee device
    Hi, I'm participating in the Bee track from India. I have installed the Bee iOS app and signed in successfully with Apple, and I have installed Bee CLI 0.7.3 on Windows.
    However, the Bee app only shows "Set up a Bee device" or "Try with Apple Watch". I don't own either device, so I cannot reach Settings/Developer Mode, and bee login reports that Developer Mode is not active.
    The hackathon FAQ says the Bee CLI shouldn't require a Bee device. Is there a supported developer/sandbox account, token, or another way for hackathon participants without a Bee device to authenticate the CLI and access test/generated Bee data?
    I want to build against the real Bee API/CLI and satisfy the Bee-track runtime requirement rather than simply mocking Bee.
    Thanks!

  •   •   8 days ago

    Hi Kourtney and Emerson,

    I'd appreciate some clarification on how judges verify that submitted data originated from a Bee device.

    I own and actively use a physical Bee wearable. I have a completed conversation and transcription in my Bee account from a recording I personally witnessed the device capturing. I've also successfully performed an authenticated, live Bee CLI read with a verified execution receipt.

    However, the conversation metadata reports device_type: "unknown". Out of curiosity, I sampled 40 additional Bee conversations across four weekly periods, and all 40 returned the same value.

    For Bee-track eligibility, is demonstrating a genuine authenticated Bee conversation/transcript, a working CLI integration, and the physical wearable sufficient?

    I want to ensure everything meets the requirements without publishing private transcripts or making unsupported hardware-attestation claims.

    Mostly, I'm just checking whether this is normal or if my device is misBEEhaving!

    Thank you!!

  •   •   6 days ago

    @Shubdeep - That FAQ line is incorrect and will be updated. You need developer mode enabled in the app to set up the CLI, and enabling developer mode DOES require pairing either an Apple Watch or Bee device. The rules also state that you must "build a project that connects to data recorded and processed through a Bee device or Apple Watch running Bee software, using Bee's CLI, MCP, or Agent Skills."
    https://amazonappdev2026.devpost.com/rules#:~:text=own%20agentic%20tools.-,Bee%20(Wearable%20AI),-%3A%20Build%20a

  •   •   6 days ago

    @Eric I'm checking with the team on that issue, but don't worry - it won't impact the judging of your submission.

Log in or sign up for Devpost to join the conversation.