MoonshotAI / MoonshotAI/kimi-code
maxRecentUserMessages配置项可能有bug。设置了该值后,可能不能正常使用
Nobody has claimed this yet.
- Dominant language
- TypeScript
- Stars
- 7.5k
- Forks
- 1.2k
- Avg merge
- 11h 53m
- Merged PRs (30d)
- 350
Description
涉及文件:packages/agent-core/src/agent/compaction/strategy.ts
涉及方法:DefaultCompactionStrategy
问题代码:
for (; recentMessages < messages.length; recentMessages++) {
const splitIndex = messages.length - recentMessages - 1;
const m2 = messages[messages.length - recentMessages]!;
if (m2.role === 'user') {
recentUserMessages++;
}
recentSize += estimateTokensForMessage(m2);
if (canSplitAfter(messages, splitIndex)) {
bestN = splitIndex + 1;
}
const reachesMax = recentMessages >= this.config.maxRecentMessages
|| recentUserMessages >= this.config.maxRecentUserMessages
|| recentSize >= this.maxSize * this.config.maxRecentSizeRatio;
if (reachesMax && bestN !== undefined) {
break;
}
}
复现过程:用消息序列 [asst_tc, tool, tool, user, user, user] 遍历时的过程:
| 轮次 | recentMsgs | m2角色 | user计数 | canSplit | bestN | reachesMax | 行为 |
|---|---|---|---|---|---|---|---|
| 1 | 1 | user | 1 | false | undefined | true(1≥1) 但 bestN=undefined | 继续 |
| 2 | 2 | user | 2 | false | undefined | true(2≥1) 但 bestN=undefined | 继续 |
| 3 | 3 | user | 3 | true | 3 | true(3≥1) 且 bestN≠undefined | break → 返回 3 |
问题:第 3 轮时 recentUserMessages 已经是 3,却错误返回了值
原因:不知道是什么问题导致,但是看起来没有什么实际影响
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start in packages/agent-core/src/agent/compaction/strategy.ts and inspect DefaultCompactionStrategy, especially the loop and maxRecentUserMessages condition. Reproduce the behavior with [asst_tc, tool, tool, user, user, user], then determine the intended split boundary and whether the returned value is valid. Done means the expected behavior is documented and covered by a regression test.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- typescript
- Domain
- ai, cli
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 45/100