Problem (one or two sentences)
If you press the Stop button while it is compressing context, the UI looks like it has stopped but it is still running the compress context request against the AI burning tokens. If you then start a new task, and your server only accepts tasks sequentially, you have to wait for that long compression task to complete before it will start the new task. Even if it does accept them in parallel, it still affected performance or cost if you pay-per-token.
_
Context (who is affected and when)
Happens to anyone who causes context compression to occur and tries to stop it.
Reproduction steps
- Make a chat session go on for long enough that automatic context compression kicks in.
- Press the Stop button in the chat window.
Expected result
The HTTP call to the AI to compress the context is aborted, and is only resumed if the user presses the continue button to resume that task
Actual result
The HTTP call to compress the context keeps running, even if you start a new task
Variations tried (optional)
I'm using llama-server. I tried killing the backend server process and restarting it but Zoo still resubmitted the HTTP request to compress the context before it submitted the new task.
App Version
3.82.0
API Provider (optional)
OpenAI Compatible
Model Used (optional)
N/A
Zoo Code Task Links (optional)
No response
Relevant logs or errors (optional)
Problem (one or two sentences)
If you press the Stop button while it is compressing context, the UI looks like it has stopped but it is still running the compress context request against the AI burning tokens. If you then start a new task, and your server only accepts tasks sequentially, you have to wait for that long compression task to complete before it will start the new task. Even if it does accept them in parallel, it still affected performance or cost if you pay-per-token.
_
Context (who is affected and when)
Happens to anyone who causes context compression to occur and tries to stop it.
Reproduction steps
Expected result
The HTTP call to the AI to compress the context is aborted, and is only resumed if the user presses the continue button to resume that task
Actual result
The HTTP call to compress the context keeps running, even if you start a new task
Variations tried (optional)
I'm using llama-server. I tried killing the backend server process and restarting it but Zoo still resubmitted the HTTP request to compress the context before it submitted the new task.
App Version
3.82.0
API Provider (optional)
OpenAI Compatible
Model Used (optional)
N/A
Zoo Code Task Links (optional)
No response
Relevant logs or errors (optional)