Every commit scored for change-risk against this repo's own history, so 'elevated' means elevated here rather than on some global curve.
Needs review
87 commits sit in this repo's top risk tercile, which is 34% of the 253scored. The cut is drawn against this codebase's own history rather than a global curve, so a quiet repo still fills its top band, and here it starts at 9.3 out of 10. What pushes a commit up is size and spread together: a large change confined to one area scores below a smaller one scattered across a dozen files.
Commit categories over time, read off the subject line. Fixes carry the accent because that is the series this chart exists to show.
Began fix-led, now leaning feature.
Ranked by change-risk, highest first. Priority is a tercile of this repo's own distribution, so a quiet repo still fills its top band.
| # | Commit | Author | When | Lines | Risk | Top driver |
|---|---|---|---|---|---|---|
| 1 | 530cd59achore(python): 补齐 kl-graph 新增依赖的 venv 产物（ladybug / omegaconf） | shenjunjiekodaclaudeassisted | 1w ago | +27.5K -1 | 99%Elevated | more lines added than baseline |
| 2 | 11eb0164refactor(kl-graph): 项目规范化 — 配置重构、模块整理、首次纳入完整代码与测试 | luckiezhou | 1w ago | +33.6K -3.7K | 99%Elevated | more lines added than baseline |
| 3 | efb97c2bfeat: initial MyContext snapshot | shenjunjiekoda | 1w ago | +6.2M -9 | 99%Elevated | more lines added than baseline |
| 4 | e766d4dffeat(kl-graph): unified incremental community detection via HIT-Leiden | Weiding.H | 3d ago | +6.5K -665 | 98%Elevated | more lines added than baseline |
| 5 | 26b0a497feat(evaluation): sync benchmark evaluation pipelines | h66840 | 3d ago | +13.2K -835 | 98%Elevated | more lines added than baseline |
| 6 | 56b44e5dfeat(distill): work 层五个 facet + playbook 归纳，修四处让结论不可信的静默失效 | Three-Colorsclaudeassisted | 6d ago | +5.0K -113 | 98%Elevated | more lines added than baseline |
| 7 | e559ae1bchore(python): 补齐 zvec 默认向量后端的 venv 产物 | shenjunjiekodaclaudeassisted | 6d ago | +9.3K -1 | 98%Elevated | more lines added than baseline |
| 8 | 02b1b218test(kl-graph): migrate tests for the GraphRAG community + search rewrite | Weiding.H | 1w ago | +3.5K -2.3K | 98%Elevated | more lines added than baseline |
| 9 | fd311b0brefactor(persona): 把回复链路拆成收/生成/管控/发四段 —— 安全管控散在七处是根因 | Three-Colors | 1w ago | +5.3K -1.5K | 98%Elevated | more lines added than baseline |
| 10 | f0b7dee4refactor(数据平面 v4): 采集面改并集 + 投递合成一条路 + 收窄告知 | shenjunjiekodaclaudeassisted | 6h ago | +4.3K -307 | 95%Elevated | more lines added than baseline |
| 11 | fc5540e8feat(recovery): 摄取恢复/自愈链路 + 断点续传迁移与评审修复 | luckiezhou | 3d ago | +2.9K -53 | 95%Elevated | more lines added than baseline |
| 12 | 247efeaerefactor(evaluation): 统一按案例编排评测流水线 | h66840 | 3d ago | +2.3K -1.2K | 95%Elevated | more lines added than baseline |
| 13 | deb953c9refactor(evaluation): align benchmark pipeline layouts | h66840 | 3d ago | +3.6K -3.0K | 95%Elevated | more lines added than baseline |
| 14 | 30435157feat(evaluation): configure LongMemEval pipeline from YAML | h66840 | 3d ago | +1.6K -680 | 95%Elevated | more lines added than baseline |
| 15 | fe06b17ffeat(community): incremental community detection with gated re-summarization | Weiding.H | 3d ago | +2.2K -72 | 95%Elevated | more lines added than baseline |
| 16 | 620506a5feat(数据平面): 路由下沉修监听范围漏网 + 拓扑可见 + 文档覆盖面(v29) | shenjunjiekoda | 4d ago | +3.4K -607 | 95%Elevated | more lines added than baseline |
| 17 | 114fd8befix(desktop): 让 work 层真的跑起来，让 agent 真的用得上 kl —— 三处静默失效 | Three-Colorsclaudeassisted | 6d ago | +3.0K -132 | 95%Elevated | more lines added than baseline |
| 18 | ae4437e9feat(kl-graph): source-processing plans, in-step extraction retry, configurable ask | luckiezhou | 6d ago | +2.3K -296 | 95%Elevated | more lines added than baseline |
| 19 | 02129c5ddocs(数据平面): v3 重构成"现状 + 判据 + 实施结果"，去掉规划语气 | shenjunjiekodaclaudeassisted | 6h ago | +1.6K -0 | 90%Elevated | more lines added than baseline |
| 20 | c06de8a2feat(数据平面): 让声明有执行力 —— 三个消费者进 runCycle、监听范围三态、域声明按渠道 | shenjunjiekodaclaudeassisted | 6h ago | +2.3K -167 | 90%Elevated | more lines added than baseline |
| 21 | 17583d30feat(数据平面): 修文档超范围采集 + 三态范围判据收敛 + 消费者声明补齐 | shenjunjiekodaclaudeassisted | 2d ago | +2.3K -170 | 90%Elevated | more lines added than baseline |
| 22 | 3005bdb5feat(ingest): 坏轮次跳过而非硬失败，stop-and-cleanup 统一到 stop | luckiezhou | 3d ago | +1.3K -512 | 90%Elevated | more lines added than baseline |
| 23 | b537e5b2feat(community): maintain communities incrementally with HIT-Leiden | Weiding.H | 3d ago | +1.8K -57 | 90%Elevated | more lines added than baseline |
| 24 | 0be71114feat(feishu): 落地飞书数据采集的初版实现（重构前的对照物） | nicoxmk | 6d ago | +2.5K -364 | 90%Elevated | more lines added than baseline |
| 25 | 2b0f3677feat(forge): signals-v6 —— 给测量加时间衰减与窗口，避免半年前的习惯与本周等权 | Three-Colorsclaudeassisted | 6d ago | +1.4K -141 | 90%Elevated | more lines added than baseline |
| 26 | 8ab6b9f1feat(dashboard): 数据趋势图 + 消化漏斗 + 数字滚动（CountUp） | shenjunjiekodaclaudeassisted | 6d ago | +2.6K -9 | 90%Elevated | more lines added than baseline |
| 27 | 4accca83feat: add source-aware extraction and entity cleanup | luckiezhou | 1w ago | +1.3K -144 | 90%Elevated | more lines added than baseline |
| 28 | c137960ffeat(kl-graph): migrate community detection and building (area 1) | Weiding.H | 1w ago | +1.2K -1.9K | 90%Elevated | more lines added than baseline |
| 29 | f762a546feat(minutes): 听记两条分页都抽干 —— 首版只取第一页且截断完全不可见 | shenjunjiekodaclaudeassisted | 1w ago | +2.1K -115 | 90%Elevated | more lines added than baseline |
| 30 | 9d9e0c69perf(kl-graph): 增量 improvement 消除三个 O(V/E) 瓶颈 | luckiezhou | 1w ago | +1.4K -143 | 90%Elevated | more lines added than baseline |
| 31 | b6f7eebefeat(kl-graph): add state-based ingestion improvement | luckiezhou | 1w ago | +1.3K -2.7K | 90%Elevated | more lines added than baseline |
| 32 | b833cd69feat(reset): 完整重置脚本 + 修对账活锁与清库后水位不归零 | shenjunjiekodaclaudeassisted | 1w ago | +2.7K -53 | 90%Elevated | more lines added than baseline |
| 33 | 8e9b4777feat(identity): 隔离维度从「本地账号」改成「渠道身份」，落点全部按 vault | shenjunjiekodaclaudeassisted | 1w ago | +2.4K -299 | 90%Elevated | more lines added than baseline |
| 34 | c6e4022efeat(数据平面 v4 阶段 D): 明细层只打标不筛行 —— 越界消息入库、消费者按标签取 | shenjunjiekodaclaudeassisted | 6h ago | +1.9K -237 | 85%Elevated | more lines added than baseline |
| 35 | 9d0a7380feat(生产者): ProducerRunner 真的接线 + 生产者运行时视图 | shenjunjiekodaclaudeassisted | 6h ago | +1.2K -253 | 85%Elevated | more lines added than baseline |
| 36 | e3463162feat(范围): per-domain 分区白名单与时间窗 —— 文档终于能只学指定知识库 | shenjunjiekodaclaudeassisted | 6h ago | +1.4K -14 | 85%Elevated | more lines added than baseline |
| 37 | e9abce3crefactor(ingest): 采集策略改成「命名策略注册表」，两轴模型收进 chat_processing | luckiezhou | 3d ago | +1.3K -233 | 85%Elevated | more lines added than baseline |
| 38 | f5e77fbbfeat(longmemeval): add configured Khoj pipeline | h66840 | 3d ago | +869 -319 | 85%Elevated | more lines added than baseline |
| 39 | 7ba69f3bfeat(渠道数据架构): 监听范围成为一等范围 + 路由模块门控投递 | shenjunjiekoda | 4d ago | +1.4K -31 | 85%Elevated | more lines added than baseline |
| 40 | 5d7a7030feat(dashboard): 问候语粒子字/大数字 CountUp/侧栏 Shuffle/图谱空态流体 + 刷新按钮进顶栏 | shenjunjiekodaclaudeassisted | 6d ago | +1.6K -122 | 85%Elevated | more lines added than baseline |
| 41 | beb3f249feat(kl-graph): unify vector storage backends | luckiezhou | 1w ago | +1.3K -301 | 85%Elevated | more lines added than baseline |
| 42 | 880dc53ffeat(kl-graph): add GraphRAG local search + engine synthesis hook (areas 3-4) | Weiding.H | 1w ago | +1.0K -8 | 85%Elevated | more lines added than baseline |
| 43 | 7000d6e0fix(kl-graph): correct incremental improvement scoping | luckiezhou | 1w ago | +984 -192 | 85%Elevated | more lines added than baseline |
| 44 | 3395a3c3feat(settings): 「重置蒸馏水位」换成「清空当前渠道的数据」 | shenjunjiekoda | 1w ago | +1.3K -31 | 85%Elevated | more lines added than baseline |
| 45 | f016cadbfix(采集): 三个真机日志暴露的 bug —— 无限重试、归因错误、不可读计数缺失 | shenjunjiekodaclaudeassisted | 6h ago | +1.0K -12 | 81%Elevated | more lines added than baseline |
| 46 | 31cbeeb3feat(生产者): 闸门判据收成一份 —— 顺带修掉飞书的时间闸失效 | shenjunjiekodaclaudeassisted | 2d ago | +1.1K -50 | 81%Elevated | more lines added than baseline |
| 47 | 5769832bfeat(学习范围): 覆盖面可见（已有多少/齐没齐）+ 把监听范围与学习范围分开 | shenjunjiekoda | 4d ago | +1.4K -10 | 81%Elevated | more lines added than baseline |
| 48 | ecbfc17ffix(飞书/头像): 两步授权在 UI 上分成三颗按钮；config remove 的成功不再被当失败；头像按渠道路由 | shenjunjiekodaclaudeassisted | 5d ago | +987 -88 | 81%Elevated | more lines added than baseline |
| 49 | c7bf0195feat(persona): migrate deterministic layer-a pipeline | luckiezhou | 5d ago | +977 -0 | 81%Elevated | more lines added than baseline |
| 50 | 18d25736feat(channels): 采集管线按已连渠道挂载，非主渠道不再是一条静默死链 | spikeeehouhua | 6d ago | +1.1K -111 | 81%Elevated | more lines added than baseline |
Two views of the same model: where the cuts fall, and what commit shape lands you above them.
Every scored commit, binned on the raw 0 to 10 score rather than the percentile. Percentile ranks are uniform by construction, so that axis has no shape to draw. The dashed lines are the tercile cuts behind each row's priority pill.
The 200 most recent commits, on their own recency sample rather than the feed above: that defaults to risk-sorted, so reusing it would plot only the top tercile and call it the spread. Big and scattered is what the model penalises. Click a dot to open it.
2233admin/mycontext has 253 commits in its history from 8 contributors, the first of them Aug 3, 2026. In the last 90 days 10,086 files were touched, 11,716 times in total, most often startup.ts. Every commit is scored for change risk from its size, spread and the history of the files it touches.
