An embedding request is sent from your local machine and may end up being processed on another Mac before the result comes back.
Desktop nodes share bge-m3 vector computation with 1024 dimensions. Your local machine exposes an OpenAI-compatible /v1/embeddings endpoint, and requests are automatically forwarded to online nodes. The fastest responder gets picked first; if it fails, the request moves on to the next node.
To provide computing power, Apple Silicon Mac users can turn on the switch on the “Computing Power” page. It uses about 1 GB of disk space, and the model is automatically downloaded and verified the first time. You can also use it without installing a client: sign with a Dogecoin address on the web portal to get an API Key—no account registration or balance checks required. Usage can be checked by address; only counts are recorded, not content.
The trade-off: text is sent in plain text to other computing nodes, so sensitive content should not be sent this way. Currently, providers receive no rewards, and there is no promise of future airdrops, points, or tokens. The local client is free; the official portal has daily limits and rate limits. See the page for details. The current stable release is v1.13.12 (released 2026-10-21) and requires macOS.
The text appears in plain text on whichever machine processes the request.
#Dogecoin #Web3 $DOGE
For technical details, refer to the official download page. This is not investment advice.