Overview
This release adds a configurable compression style for summary generation. Operators can now choose whether summaries should prioritize maximum token savings, the default balanced continuity, or higher-fidelity preservation of nuanced context.
New Features
compression_stylevalve: Addedaggressive,balanced, andfaithfulmodes.- Aggressive mode: Minimizes token use by keeping only the facts most likely to affect the next reply.
- Balanced mode: Keeps the previous default tradeoff, preserving decisive facts, key rationale, active alternatives, constraints, and unresolved questions while removing low-value detail.
- Faithful mode: Uses more of the configured summary budget to preserve reasoning chains, evaluation criteria, candidate options, materially distinct facts, and examples that clarify why context matters.
Configuration Notes
compression_style defaults to balanced, so existing installs keep the same practical behavior unless the valve is changed.
Use:
aggressivewhen context cost is the main concern.balancedfor general long-chat continuity.faithfulfor coding, planning, analysis, and review threads where losing nuance can change future answers.
Migration Notes
No breaking changes. Update or reinstall the filter so OpenWebUI's stored function content includes the new valve metadata and prompt instructions.
Version Changes
Plugin Updates
- Async Context Compression: v1.6.4 → v1.6.5 | 📖 README
New Contributors
📚 Documentation Portal
🐛 Report Issues
Full Changelog: async-context-compression-v1.6.4...async-context-compression-v1.6.5