Configure context compression per session
- Dominant language
- Python
- Stars
- 21.5k
- Forks
- 4k
- Avg merge
- 1d 14h
- Merged PRs (30d)
- 37
Description
**Is your feature request related to a problem? Please describe.**
Currently, context compression is defined on an `App` level. So every session on that app will share the same compression config (including on/off).
I want to control whether compression is turned on/off on a per session bases, based on the user/other metadata.
**Describe the solution you'd like**
Maybe have a lifecycle callback on the `App` layer to configure these options. Something `before_session_start`.
**Describe alternatives you've considered**
Create my own compaction events, using `after_agent` callback.
**Additional context**
I am deploying using Vertex AI AgentEngine, so want to be aligned with the possibilities there.
Contributor guide
Assessment
This issue has not been assessed yet.