Conversation
Add SessionPrompt.continue() and POST /:sessionID/continue that re-enters the agentic loop after an interruption without creating a new user message. Preserves all completed thinking, text and tool results (abort cleanup already finalized them). Ref anomalyco#7226, Ref anomalyco#5150, Ref anomalyco#8197, Ref anomalyco#10374, Ref anomalyco#6653, Ref anomalyco#19023
github-actions
bot
added
needs:compliance
|
Thanks for updating your PR! It now meets our contributing guidelines. 👍 |
|
I've had my own take on this since early December at ariane-emory#11 that's been working very well for me. Recommend adding a |
Thanks for sharing your PR! Didn't know that there was one open still. Its basically the same as mine, even superior in some aspects. My idea was for /continue to really pick up the session as if no interruption had ever happend - picking up thinking where it stopped half-way. Assistant prefill could be a lever to make this happen. But as it seems Anthropic has deprecated assistant prefill recently and if I am not mistaken the best we can get is something that writes a user message "Your session got interrupted, please continue".... That TBH doesn't justify a built-in core feature :(, as it can be already done with what we have. Unless someone has other ideas? Closing the PR for now |
2 participants
Issue for this PR
Closes #7226, #5150, #8197, #10374, #6653, #19023
Type of change
What does this PR do?
Adds a /continue command so the user can call /continue if a session got interrupted (for example by a network dropout) without the need to type a new user message and waste tokens or distract the model in its flow
If you paste a large clearly AI generated description here your PR may be IGNORED or CLOSED!
How did you verify your code works?
Screenshots / recordings
Checklist
If you do not follow this template your PR will be automatically rejected.