fix(session): 剔除 Session 标题中的 System Instruction 噪声#246
Merged
ThreeFish-AI merged 1 commit intoMay 25, 2026
Conversation
Claude Code 会在首条 user 消息 content 中拼接多个 <system-reminder> / <user-preferences> 等系统注入块,导致原 _extract_session_title 抽取出的 标题在不同会话间高度同质,丧失辨识度。 - 新增 _sanitize_user_text,基于白名单剥离 system-reminder、user-preferences、 local-command-stdout/stderr、bash-input/stdout/stderr、ide_selection、 stdin、system_instruction 等 CC 注入标签,折叠空白后返回真实用户输入; - 优先识别 slash command (<command-name>/<command-args>),合成"命令 + 参数" 式标题,避免命令式会话标题为空; - 重写 _extract_session_title,跳过清洗后为空的 user 文本 part,自动 fallback 到下一个有真实内容的 part; - 补充 20 个单元测试覆盖噪声剥离、slash command、空白折叠、截断、边界场景。 🤖 Generated with [Claude Code](https://github.com/claude), [CodeX](https://openai.com), [Gemini](https://github.com/apps/gemini-code-assist) Co-Authored-By: Aurelius Huang<threefish.ai@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
背景
PR #245 已实现「从首个用户消息提取 Session 标题并在 Dashboard 展示」，但
_extract_session_title直接取首个 user 文本块前 30 字，未识别 Claude Code 在首条 user 消息 content 中拼接的<system-reminder>/<user-preferences>等系统注入块，导致 Dashboard?tab=sessions中所有 Session 标题都退化为同质的「MCP Server Instructions...」模板段，丧失辨识度。变更要点
src/coding/proxy/routing/executor.py：_sanitize_user_text，基于白名单剥离 CC 注入标签 (system-reminder、user-preferences、local-command-stdout/stderr、bash-input/stdout/stderr、ide_selection、stdin、system_instruction)，正则使用re.DOTALL | re.IGNORECASE并容忍标签携带属性 (<tag attr="x">);<command-name>时合成「命令 + 参数」式标题 (例/commit feat: 新增标题清洗)，避免命令式会话标题为空；<command-*>包裹标签，最终折叠连续空白；_extract_session_title：对清洗后为空的 user 文本 part 自动 fallback 到下一个非空 part。tests/test_router_executor.py：新增TestSanitizeUserText(12 例) 与TestExtractSessionTitle(8 例) 共 20 个单元测试，覆盖噪声剥离、多行 / 带属性标签、slash command、空白折叠、30 字截断、跳过 assistant 角色、保留用户合法 XML 等场景。设计取舍
<div>、Markdown 代码块等内容；routing/executor.py而非compat/canonical.py解析阶段，保持代理透明性 (vendor 转发请求体不受影响)，仅在生成 session 标题时应用；session_meta表使用INSERT OR IGNORE，历史数据无需迁移，老记录标题原样保留，仅新建 Session 写入清洗后标题。验证
uv run pytest tests/test_router_executor.py— 121/121 通过 (101 既有 + 20 新增)；uv run pre-commit run --files src/coding/proxy/routing/executor.py tests/test_router_executor.py— 全绿。🤖 Generated with Claude Code