google / google/adk-python

Enabling explicit context caching alongside context filtering

Open
#3,231 2 comments 0 reactions 2 assignees Claimed by @klateefa View on GitHub
core needs review
Dominant language
Python
Stars
21.5k
Forks
4k
Avg merge
1d 14h
Merged PRs (30d)
37

Description

When explicit context caching is enabled concurrently with context filtering for a maximum of N contents, any new content added beyond the N limit will trigger a cache cleanup and the creation of a new cache without any reusing. This results in offering no performance benefit but incurring an unnecessary cost overhead.

**Request**: We need to add a configuration option to Explicit Context Caching that allows caching to be limited exclusively to static context elements (e.g., system instructions and tool descriptions).
**Justification**: In many designs, the required conversational history is inherently short, effectively constrained by Context Filtering to N recent items. However, the static instructions and knowledge base can be quite large. By caching only the long static components, we prevent the costly cache regeneration/thrashing that occurs when dynamic content (history) exceeds the N limit. This provides the performance benefit of caching the static overhead without the cost of repeatedly invalidating a full cache for short histories.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.