Feature Request: Configurable compaction threshold percentage

[smnuman]

Description

It would be great to have a configurable threshold for when auto-compaction triggers, specified as a percentage of context window usage.

Use Case

Instead of compaction only triggering when context is ~100% full, users should be able to set a percentage (e.g., 50-62%) to proactively compact before hitting limits.

Proposed Solution

Add a new config option like:

{
  \"compaction\": {
    \"auto\": true,
    \"threshold\": 0.5,  // 50% of context window
    \"prune\": true,
    \"reserved\": 10000
  }
}

Or alternatively, a percentage string:

{
  \"compaction\": {
    \"threshold\": \"50%\"
  }
}

Current Workaround

Currently users can set a high reserved value to force earlier compaction, but this isn't precise and doesn't achieve the same goal.

Context

• Model: antigravity-claude-opus-4-6-thinking (200k context)
• User request: trigger compaction at 50-62% token usage

Thanks for considering this!

[github-actions]

This issue doesn't fully meet our contributing guidelines.

What needs to be fixed:

• Feature request titles must start with [FEATURE]: (current title uses "Feature Request:" instead)
• Feature request template requires a verification checkbox confirming the feature doesn't already exist

Please edit this issue to address the above within 2 hours, or it will be automatically closed.

---

This issue might be a duplicate of existing issues. Please check:

• Feature Request: Configurable Context Compaction Threshold #11314: Feature Request: Configurable Context Compaction Threshold
• Feature Request: Custom compaction threshold to trigger earlier #10017: Feature Request: Custom compaction threshold to trigger earlier
• Feature Request: Configurable context limit and auto-compaction threshold #8140: Feature Request: Configurable context limit and auto-compaction threshold
• Feature: Configurable compaction threshold and model (global + per-model) #11930: Feature: Configurable compaction threshold and model (global + per-model)

If you believe this was flagged incorrectly, please let a maintainer know.

[github-actions]

This issue has been automatically closed because it was not updated to meet our contributing guidelines within the 2-hour window.

Feel free to open a new issue that follows our issue templates.