Repository navigation
fix(prompts): 自然闲聊与侧写事实判断（v3.18.8） - #113
Conversation
Distinguish casual recognition from requests for implementation details, and avoid unasked system explanations and defensive caveats.
Make spaces the default replacement for casual commas while preserving required punctuation in formal prose, quotations, code, and data.
Require contextual evidence before treating claims as identity facts across observation collection, rewriting, and profile merging. Do not count repetition as corroboration, allow missing profiles to remain absent, and remove confirmed misrecords without inventing replacement identities.
Summarize natural conversation and profile fact handling changes, and synchronize the backend and native client versions.
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
|
No actionable comments were generated in the recent review.
 |
| Layer / File(s) | Summary |
|---|---|
Conversational response guidance docs/usage.md, res/IMPORTANT/each.md, res/prompts/undefined*.xml, tests/test_prompt_structure.py |
Prompt and documentation guidance now favors space-separated pauses in casual short messages, limits unsolicited explanations of memory, and distinguishes casual remarks from direct questions about assistant capabilities. |
Observation evidence rules res/IMPORTANT/each.md, res/prompts/undefined*.xml, src/Undefined/skills/tools/end/config.json, tests/test_prompt_structure.py |
Guidance distinguishes reliable self-reports and corrections from jokes, repetition, role-play, and unconfirmed claims. Historical context may clarify statements but cannot supply new observation facts. Tests check shared guidance and memory regression cases. |
Historian rewrite and profile merging docs/cognitive-memory.md, docs/usage.md, res/prompts/historian_*.md, src/Undefined/cognitive/historian/tools.py, tests/test_cognitive_historian.py, tests/test_prompt_structure.py |
Historian guidance checks source context and reliability before recording or merging facts. It updates profile skip and correction conditions, and tests cover varied source messages, observations, and prompt evidence placeholders. |
Release version alignment CHANGELOG.md, pyproject.toml, src/Undefined/__init__.py, apps/undefined-chat/..., apps/undefined-console/... |
The changelog and project and application metadata change to version 3.18.8. |
Priority:
Estimated code review effort: 3 (Moderate) | ~20 minutes
Change: Bug fix
Possibly related PRs
- 69gg/Undefined#112: Establishes earlier casual-reply and punctuation rules in the same prompt and documentation files.
- 69gg/Undefined#111: Establishes earlier memory and task-context boundaries that this change extends to observation and profile merging.
Merge Risk: 
81118
-
This release adjusts casual-reply tone, punctuation defaults, and how memory and profiles treat jokes and repeated claims. No blocking risk was found. Prompt behavior has not been tested against real models, and existing profiles are not cleaned automatically, but that is documented and intended.
Pre-merge checks: 
-
Check name Status Explanation Title check 标题明确说明了自然闲聊、侧写事实判断和版本变更，准确覆盖了本次 PR 的主要改动。 Description check 描述完整覆盖变更说明、影响范围、关联 Issue、自检和备注。已说明测试结果、限制条件、部署注意事项，以及配置项和界面截图不适用的原因。 Docstring Coverage Docstring coverage is 80.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 10 functions across 4 files. (16 skipped: 1… Linked Issues check Check skipped because no linked issues were found for this pull request. Out of Scope Changes check Check skipped because no linked issues were found for this pull request. Full details: Docstring Coverage
-
Explanation
Docstring coverage is 80.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 10 functions across 4 files. (16 skipped: 16 unsupported.)
-
Finishing Touches
-
📝 Generate docstrings
-
- Commit to this branch
- Create a new PR
🧪 Generate unit tests (beta)
-
- Commit to this branch
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Comment @coderabbitai help to get the list of available commands.
变更说明
日常随口问“你怎么还记得”时，Bot 容易主动展开记忆机制、附加没问的解释与辩解；群聊里的玩梗和复读也可能被当成真实身份。本 PR 调整提示词中的表达取舍和事实判断，并统一版本为
3.18.8，补充对应 changelog。影响范围
3.18.8，依赖版本保持一致。关联 Issue
无。
自检
uv run ruff check .、uv run ruff format --check .与uv run mypy .（提交钩子执行）UNDEFINED_DEPLOY_WHEEL_CHECK=1 uv run pytest tests/ --cov --cov-report=term -q（4611 项通过、1 项跳过，覆盖率 70.27%，达到 65% 门槛）npm --prefix apps/undefined-console run check（提交钩子执行）npm --prefix apps/undefined-chat run check（提交钩子执行，含 507 项单测、50 项端到端测试、构建回归及 76 项 Rust 测试）uv run python scripts/release_notes.py validate --tag v3.18.8uv build --wheel，并核对打包后的提示词、end 工具说明和版本文件与源码一致备注
提示词调整尚未做真实模型对照，不能保证识破所有玩笑或误导。既有侧写不会自动批量清洗，后续合并取得明确纠错依据时才修正。更新随包提示词后需重启 Bot 加载资源。
Summary by CodeRabbit
对话体验
记忆与侧写