Avaer Kazmer
Avaer Kazmer
### What do you want to see? Allow agents to run on the desktop using [`electron`](https://github.com/electron/electron). Automation courtesy of [@nut-tree-fork/nut-js](https://www.npmjs.com/package/@nut-tree-fork/nut-js). This is to support a live desktop control experience similar...
We should use either heuristics or `mini` prepass for detecting whether the agent should reply to a given perception. We can use `` for this.
### What do you want to see? We need to add the cost functions to the proxy. ### Link to original discussion _No response_
Repro: - `/login` - https://login.upstreet.ai/logintool?callback_url=https://example.com Result:
We already support uploading custom voices for agents. The user does this with the `usdk voice` command which is now unlocked. We need to document + tutorialize the flow of...
We should queue the audio perceptions better so the agent waits for their turn to speak.
In Discord voice we were relying on the client start/stops, but they are too noisy. If we merge the streams per user (with inserted silence) and use our own VAD...
`usdk` can be imported as a module to get access to the cli functionality. The commands are already well-factored and the CLI is just a wrapper around this so the...
Agents have conversational + embedded memory + kv store. Sometimes we want to clear the memory, for debugging or other purposes. For this we should introduce a `usdk reset` or...
We have a ready implementation for reading https://github.com/malfoyslastname/character-card-spec-v2 Many character communities use these: https://character-tavern.com/ https://chub.ai/ We can add this to the agent creation pipeline e.g. ``` usdk create -i ./liko.png...