Skip to content
TheChatArchiveThe Chat Archive
composer-2.5· unknownHugging Face ibragim-bad/swe_rebench_07_2026_trajectories

Bug: retainrecentlegalsuffix 在 else 分支下会导致用户消息被重复归档 Bug Description 在…

  1. User

    Bug: retain_recent_legal_suffix 在 else 分支下会导致用户消息被重复归档

    Bug Description

    在 nanobot/session/manager.py 的 retain_recent_legal_suffix 方法中,当 else 分支(最近 N 条消息全是助手/工具消息)被触发时,由于 cut 值的计算使用 len(tail) - len(kept),而没有考虑用户消息的实际位置,导致用户消息被错误地同时归档到 archive 和 kept 中。

    具体影响:

    • 用户消息同时出现在 archive(待压缩)和 kept(保留)中
    • 可能导致 LLM 上下文不一致

    Steps to Reproduce

    1. 会话中消息数量超过 _RECENT_SUFFIX_MESSAGES (默认10条)
    2. 最近 N 条消息中 不包含任何用户消息 (全是 assistant/tool 角色)
    3. 在完整会话历史中,最后一条用户消息位于倒数第 11-15 条之间

    Expected Behavior

    当会话历史中最近 N 条消息全是助手/工具消息,而用户消息位于更早位置时:

    1. 用户消息不应被重复归档 - 用户消息应该只保留在 kept 中,不应该同时出现在 archive 和 kept 中
    2. 归档边界应该正确 - cut 值应该基于用户消息的实际位置,而不是简单的 len(tail) - len(kept)
    3. 上下文连贯性应该保持 - 从用户消息开始的完整上下文应该被保留,不应该被错误切割

    假设:

    tail = messages[0:100](100条)

    latest_user = 85(用户消息位置)

    max_messages = 10

    期望:

    cut = latest_user # 应该是 85,不是 90 archive = tail[:85] # [0-84],不包含用户消息 kept = tail[85:] # [85-99],包含用户消息及其后续内容

    结果:

    archive 和 kept 没有重叠 ✅

    用户消息只在 kept 中 ✅

    上下文连贯 ✅

    总结

    期望行为 :无论用户消息位于何处,它都应该:

    1. 只出现在 kept 中(保留消息)
    2. 不出现在 archive 中(归档消息)
    3. 作为保留消息的起始点,确保上下文连贯

    Relevant Logs

    nanobot Version

    0.2.0

    Python Version

    3.11

    Operating System

    Windows

    Channel / Platform

    Weixin (Personal WeChat)

    LLM Provider

    OpenAI

    Configuration (Optional)

    Additional Context

    No response

    Interface

    Type: Function Name: retain_recent_legal_suffix Path: nanobot/session/manager.py Input: self, max_messages: int Output: tuple[list[dict], int] Description: Trims the session's message list to at most max_messages while preserving a legal suffix (ending with a user message if possible). Returns the list of dropped messages and a count of how many of those were already in the consolidated prefix, so callers can archive only new, unconsolidated messages.

  2. Assistant

    Investigating retain_recent_legal_suffix to locate the duplicate-archive bug and fix the cut calculation.

  3. Assistant

    The current implementation differs from the bug report. Searching for related logic and tests.

  4. Assistant

    Confirming the bug: callers compute cut = len(tail) - len(kept), which fails when the else branch anchors to an earlier user message. Implementing the fix.

  5. Assistant

    Adding a test for the duplicate-archive scenario:

  6. Assistant

    Summary

    The bug was in how the archive boundary was computed when retain_recent_legal_suffix anchored to an earlier user message (else branch: recent N messages are all assistant/tool).

    retain_recent_legal_suffix correctly kept messages starting at latest_user, but callers in memory.py and test_auto_compact.py computed:

    cut = len(tail) - len(kept)
    

    That assumes kept is a suffix of tail. When the user message is outside the recent tail (e.g. at index 85 in a 100-message tail with max_messages=10), cut became 90 instead of 85, so the user message landed in both archive and kept.

    Fix

    1. retain_recent_legal_suffix now tracks keep_start through all trimming steps (user anchoring, legal tool boundary, hard cap) and returns that index.
    2. memory.py and test_auto_compact.py use the returned cut: archive_msgs = tail[:cut].

    With your example (tail=100, latest_user=85, max_messages=10):

    • cut = 85
    • archive = tail[:85] — no user message
    • kept = tail[85:] — user message and following context only

    A regression test was added: test_retain_recent_legal_suffix_cut_anchors_to_latest_user_outside_tail.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.