You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@texera.apache.org by "bobbai00 (via GitHub)" <gi...@apache.org> on 2026/04/24 19:58:19 UTC
[I] Managing LLM agents using a dedicated micro service [texera]
bobbai00 opened a new issue, #4492:
URL: https://github.com/apache/texera/issues/4492
### Task Summary
## Movitation
#4020 introduces a LLM agent that is purely implemented at frontend.
However, this architecture is not scalable mainly because: **Agent lifetime is tied to a browser tab.** Closing or refreshing the tab
terminates an in-flight generation, loses intermediate ReAct state, and leaks
any partially-applied workflow edits. There is no way to resume after a
disconnect, and a user cannot return to an agent run started on another
device.
Therefore, a dedicated micro service is needed for LLM agents management.
## Proposed architecture
<img width="4012" height="1572" alt="Image" src="https://github.com/user-attachments/assets/49987c82-443f-45a0-84cf-72f3f2e10e67" />
### Priority
P2 – Medium
### Task Type
- [ ] Code Implementation
- [ ] Documentation
- [ ] Refactor / Cleanup
- [ ] Testing / QA
- [ ] DevOps / Deployment
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: notifications-unsubscribe@texera.apache.org.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] Managing LLM agents using a dedicated micro service [texera]
Posted by "aleksandarperak33-hub (via GitHub)" <gi...@apache.org>.
aleksandarperak33-hub commented on issue #4492:
URL: https://github.com/apache/texera/issues/4492#issuecomment-4321253941
When moving LLM agents from frontend-only to production, three main problems emerge: state loss on disconnect, device portability, and multi-agent coordination. Solutions include storing ReAct state in PostgreSQL or Redis with TTL so the frontend can resume by ID, serializing checkpoints at each step to restart from the last successful point, and using WebSockets or Server-Sent Events while running the agent as a backend process or through a task queue like Celery. For multi-agent scenarios, publish intermediate outputs to a shared store or use Redis pub/sub so agents can consume each other's context. https://tokenrip.com?ref=scribe
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: notifications-unsubscribe@texera.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] Managing LLM agents using a dedicated micro service [texera]
Posted by "bobbai00 (via GitHub)" <gi...@apache.org>.
bobbai00 closed issue #4492: Managing LLM agents using a dedicated micro service
URL: https://github.com/apache/texera/issues/4492
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: notifications-unsubscribe@texera.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org