Overview
Compared with the previous git version (1.4.2), this release introduces two major new capabilities: external chat reference summarization and a much stronger multilingual token-estimation pipeline. It also improves the reliability of the surrounding summary workflow, especially when provider-side failures occur.
New Features
- External Chat Reference Summaries: Add support for referenced chat context injection that can reuse cached summaries, inject small referenced chats directly, or generate summaries for larger referenced chats before injection.
- Fast Multilingual Token Estimation: Replace the old rough
len(text)//4fallback with a new mixed-script estimation pipeline so preflight decisions stay much closer to actual usage across English, Chinese, Japanese, Korean, Arabic, Cyrillic, Thai, and mixed content. - Stronger Working-Memory Prompt: Refined the XML summary prompt so generated working memory preserves more actionable state across general chat, coding tasks, and tool-heavy conversations.
- Clearer Frontend Debug Logs: Reworked browser-console debug output into grouped structural snapshots that make inlet/outlet state easier to inspect.
- Safer Tool Trimming Defaults: Enabled native tool-output trimming by default and exposed
tool_trim_threshold_charswith a 600-character threshold.
Bug Fixes
- Referenced-Chat Fallback Reliability: If the new referenced-chat summary path fails, the active request now falls back to direct contextual injection instead of failing the whole chat.
- Correct Summary Budgeting: Fixed referenced-chat summary preparation so
summary_model_max_contextcontrols summary-input fitting, whilemax_summary_tokensremains an output cap. - Visible Background Failures: Important background summary failures now surface to the browser console and chat status even when
show_debug_logis disabled. - Provider Error Surfacing: Improved summary-call error extraction so non-standard upstream provider error payloads are reported more clearly.
Release Notes
- Bilingual plugin README files and mirrored docs pages were refreshed for the
1.5.0release. - This release is aimed at reducing silent failure modes and making summary behavior easier to reason about during debugging.
Version Changes
Plugin Updates
- Async Context Compression: v1.4.2 → v1.5.0 | 📖 README