Skip to content

Hard threshold is identical to the auto threshold #4945

@rcannizzaro

Description

@rcannizzaro

What happened?

When I check /context I can see that the Auto and Hard tiers are identical for me. This means that compaction doesn't happen until the last possible moment.

Context Window: 60k
Effective Window: 40k
Warn: 36k
Auto: 42k
Hard: 42k

Image

What did you expect to happen?

Auto and Hard windows are not identical.

Client information

Client Information

Run qwen to enter the interactive CLI, then run the /about command.

$ qwen /about
Qwen Code v0.17.1
Model: local
Fast Model: not set
Auth: openai
Platform: linux x64 (7.0.11-1-cachyos)
Node.js: v26.1.0
Session: 779308ee-5c82-45e6-a39a-c86b371d6f14
Git commit: 09d0faf6b
LSP: disabled

Login information

No response

Anything else we need to know?

No response

Metadata

Metadata

Assignees

Labels

category/coreCore engine and logicpriority/P2Medium - Moderately impactful, noticeable problemscope/token-managementToken handling and limitstype/bugSomething isn't working as expectedwelcome-pr

Type

No type
No fields configured for issues without a type.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions