fix: split long qqofficial messages by 4000-char limit

Author: wcqqq1214

astrbot/core/platform/sources/qqofficial/qqofficial_message_event.py

@@ -4,6 +4,7 @@
 import logging
 import os
 import random
+import re
 from typing import cast
 
 import aiofiles
@@ -211,6 +212,59 @@ def _extract_response_message_id(ret) -> str | None:
         ret_id = getattr(ret, "id", None)
         return str(ret_id) if ret_id is not None else None
 
+    # QQ 官方单条消息文本长度上限，超出会被平台截断（错误码 40054007）
+    QQ_MAX_LENGTH = 4000
+
+    _SPLIT_PATTERNS = {
+        "paragraph": re.compile(r"\n\n"),
+        "line": re.compile(r"\n"),
+        "sentence": re.compile(r"[.!?。！？]"),
+        "word": re.compile(r"\s"),
+    }
+
+    @classmethod
+    def _split_message(cls, text: str) -> list[str]:
+        """按段落 → 行 → 句子 → 词回退切分超长文本，对齐 Telegram 的 _split_message。"""
+        if len(text) <= cls.QQ_MAX_LENGTH:
+            return [text]
+
+        chunks = []
+        remaining = text
+        while remaining:
+            if len(remaining) <= cls.QQ_MAX_LENGTH:
+                chunks.append(remaining)
+                break
+
+            split_point = cls.QQ_MAX_LENGTH
+            segment = remaining[: cls.QQ_MAX_LENGTH]
+            for _, pattern in cls._SPLIT_PATTERNS.items():
+                if matches := list(pattern.finditer(segment)):
+                    split_point = matches[-1].end()
+                    break
+
+            chunks.append(remaining[:split_point])
+            remaining = remaining[split_point:].lstrip()
+
+        return chunks
+
+    @classmethod
+    def _split_message_chain_by_length(
+        cls, message_chains: list[MessageChain]
+    ) -> list[MessageChain]:
+        """按 QQ 单条消息长度限制拆分消息链，防止长文本被平台截断。"""
+        result: list[MessageChain] = []
+        for chain in message_chains:
+            text = "".join(c.text for c in chain.chain if isinstance(c, Plain))
+            if len(text) <= cls.QQ_MAX_LENGTH:
+                result.append(chain)
+                continue
+            parts = cls._split_message(text)
+            non_plain = [c for c in chain.chain if not isinstance(c, Plain)]
+            # 首个分片保留非 Plain 组件（如图片），保证含媒体消息的超长文本也被拆分
+            result.append(chain.derive([*non_plain, Plain(parts[0])]))
+            result.extend(chain.derive([Plain(part)]) for part in parts[1:])
+        return result
+
     @staticmethod
     def _split_message_chain_by_media(message: MessageChain) -> list[MessageChain]:
         chunks: list[MessageChain] = []
@@ -249,11 +303,15 @@ async def _post_send(self, stream: dict | None = None):
             return None
 
         message_chains = self._split_message_chain_by_media(self.send_buffer)
-        stream_for_chain = stream if len(message_chains) == 1 else None
+        message_chains = self._split_message_chain_by_length(message_chains)
 
         ret = None
-        for message_chain in message_chains:
-            ret = await self._post_send_one(message_chain, stream_for_chain)
+        # C2C 流式一次调用是同一消息的连续分片；buffer 被拆成多段时，
+        # 只有最后一段携带 stream 载荷（保证流会话 id 连续、最终 state=10 正常结束），
+        # 其余段降级为非流式发送。
+        for index, message_chain in enumerate(message_chains):
+            chain_stream = stream if index == len(message_chains) - 1 else None
+            ret = await self._post_send_one(message_chain, chain_stream)
 
         self.send_buffer = None
 

astrbot/core/platform/sources/qqofficial/qqofficial_platform_adapter.py

@@ -336,6 +336,9 @@ async def _send_by_session_common(
         message_chains = QQOfficialMessageEvent._split_message_chain_by_media(
             message_chain
         )
+        message_chains = QQOfficialMessageEvent._split_message_chain_by_length(
+            message_chains
+        )
         if len(message_chains) > 1:
             for split_message_chain in message_chains:
                 await self._send_by_session_common(session, split_message_chain)

tests/test_qqofficial_message_length_split.py

@@ -0,0 +1,226 @@
+"""Tests for QQ Official single-message length splitting.
+
+QQ 官方 API 对单条消息文本有长度上限（约 4000 字符，超限返回错误码
+40054007）。适配器在发送前按该限制切分，防止长文本被平台截断。
+"""
+
+from __future__ import annotations
+
+import asyncio
+from types import SimpleNamespace
+from unittest.mock import AsyncMock
+
+import botpy.message
+import pytest
+
+from astrbot.api.event import MessageChain
+from astrbot.api.message_components import Image, Plain
+from astrbot.api.platform import (
+    AstrBotMessage,
+    MessageMember,
+    MessageType,
+    PlatformMetadata,
+)
+from astrbot.core.platform.message_session import MessageSession
+from astrbot.core.platform.sources.qqofficial.qqofficial_message_event import (
+    QQOfficialMessageEvent,
+)
+from astrbot.core.platform.sources.qqofficial.qqofficial_platform_adapter import (
+    QQOfficialPlatformAdapter,
+)
+
+
+def _extract_send_text(kwargs: dict) -> str:
+    text = kwargs.get("content")
+    if text:
+        return str(text)
+    md = kwargs.get("markdown")
+    if isinstance(md, dict):
+        return str(md.get("content") or "")
+    if md is not None:
+        return str(getattr(md, "content", None) or "")
+    return ""
+
+
+def _make_group_event() -> QQOfficialMessageEvent:
+    raw = botpy.message.GroupMessage(
+        api=None,
+        event_id="event-1",
+        data={
+            "id": "msg-1",
+            "author": {"member_openid": "member-1"},
+            "group_openid": "group-1",
+            "content": "ping",
+            "timestamp": "0",
+        },
+    )
+    abm = AstrBotMessage()
+    abm.message_id = "msg-1"
+    abm.session_id = "group-1"
+    abm.group_id = "group-1"
+    abm.self_id = "bot-1"
+    abm.sender = MessageMember(user_id="member-1", nickname="u")
+    abm.type = MessageType.GROUP_MESSAGE
+    abm.message_str = "ping"
+    abm.message = []
+    abm.raw_message = raw
+    meta = PlatformMetadata(name="qq_official", description="t", id="qq_official")
+    bot = SimpleNamespace(api=SimpleNamespace(post_group_message=AsyncMock()))
+    return QQOfficialMessageEvent(
+        message_str="ping",
+        message_obj=abm,
+        platform_meta=meta,
+        session_id="group-1",
+        bot=bot,  # type: ignore[arg-type]
+    )
+
+
+def _make_c2c_event() -> QQOfficialMessageEvent:
+    raw = botpy.message.C2CMessage(
+        api=None,
+        event_id="event-2",
+        data={
+            "id": "msg-c2c",
+            "author": {"user_openid": "user-1"},
+            "content": "ping",
+            "timestamp": "0",
+        },
+    )
+    abm = AstrBotMessage()
+    abm.message_id = "msg-c2c"
+    abm.session_id = "user-1"
+    abm.self_id = "bot-1"
+    abm.sender = MessageMember(user_id="user-1", nickname="u")
+    abm.type = MessageType.FRIEND_MESSAGE
+    abm.message_str = "ping"
+    abm.message = []
+    abm.raw_message = raw
+    meta = PlatformMetadata(name="qq_official", description="t", id="qq_official")
+    bot = SimpleNamespace(api=SimpleNamespace(post_group_message=AsyncMock()))
+    return QQOfficialMessageEvent(
+        message_str="ping",
+        message_obj=abm,
+        platform_meta=meta,
+        session_id="user-1",
+        bot=bot,  # type: ignore[arg-type]
+    )
+
+
+def test_split_message_respects_limit() -> None:
+    long_text = "不稀罕。" * 1500
+    chunks = QQOfficialMessageEvent._split_message(long_text)
+    assert len(chunks) > 1
+    assert all(len(c) <= QQOfficialMessageEvent.QQ_MAX_LENGTH for c in chunks)
+    assert "".join(chunks) == long_text
+
+
+def test_split_message_short_unchanged() -> None:
+    text = "短消息"
+    assert QQOfficialMessageEvent._split_message(text) == [text]
+
+
+def test_split_message_chain_by_length_keeps_short_media_chain() -> None:
+    chain = MessageChain(chain=[Plain("标题"), Image(file="x.png")])
+    assert QQOfficialMessageEvent._split_message_chain_by_length([chain]) == [chain]
+
+
+def test_split_message_chain_by_length_splits_media_caption() -> None:
+    caption = "标题" + "长" * 5000
+    chain = MessageChain(chain=[Plain(caption), Image(file="x.png")])
+    result = QQOfficialMessageEvent._split_message_chain_by_length([chain])
+    assert len(result) > 1
+    # 媒体保留在首个分片，其余分片为纯文本
+    assert isinstance(result[0].chain[0], Image)
+    assert all(not isinstance(c, Image) for c in result[0].chain[1:]) and all(
+        not any(isinstance(c, Image) for c in ch.chain) for ch in result[1:]
+    )
+    texts = ["".join(c.text for c in ch.chain if isinstance(c, Plain)) for ch in result]
+    assert all(len(t) <= QQOfficialMessageEvent.QQ_MAX_LENGTH for t in texts)
+    assert "".join(texts) == caption
+
+
+def test_split_message_chain_by_length_splits_long_text() -> None:
+    chain = MessageChain(chain=[Plain("不稀罕。" * 1500)])
+    result = QQOfficialMessageEvent._split_message_chain_by_length([chain])
+    assert len(result) > 1
+    assert all(
+        len(c.chain[0].text) <= QQOfficialMessageEvent.QQ_MAX_LENGTH for c in result
+    )
+    assert "".join(c.chain[0].text for c in result) == chain.chain[0].text
+
+
+@pytest.mark.asyncio
+async def test_post_send_splits_long_reply_into_multiple_messages() -> None:
+    event = _make_group_event()
+    captured: list[str] = []
+
+    async def capture(**kwargs):
+        captured.append(_extract_send_text(kwargs))
+        return {"id": f"out-{len(captured)}"}
+
+    event.bot.api.post_group_message = AsyncMock(side_effect=capture)
+
+    long_text = "不稀罕。" * 1500
+    await event.send(MessageChain(chain=[Plain(long_text)]))
+
+    assert len(captured) > 1
+    assert all(len(t) <= QQOfficialMessageEvent.QQ_MAX_LENGTH for t in captured)
+    assert "".join(captured) == long_text
+
+
+@pytest.mark.asyncio
+async def test_post_send_c2c_stream_split_streams_only_last_chunk() -> None:
+    event = _make_c2c_event()
+    event.send_buffer = MessageChain(chain=[Plain("不稀罕。" * 1500)])
+    sent: list[dict | None] = []
+
+    async def fake_post_c2c_message(openid, **kwargs):
+        sent.append(kwargs.get("stream"))
+        return SimpleNamespace(id=f"c2c-{len(sent)}")
+
+    event.post_c2c_message = AsyncMock(  # type: ignore[method-assign]
+        side_effect=fake_post_c2c_message
+    )
+
+    stream_payload = {"state": 1, "id": "prev-1", "index": 3, "reset": False}
+    await event._post_send(stream=stream_payload)
+
+    # 一次流式 flush 超长被拆成多段时，只有最后一段携带 stream 载荷，
+    # 保证 C2C 流会话 id 连续、最终 state=10 能正常结束；其余段非流式发送。
+    assert len(sent) > 1
+    assert all(s is None for s in sent[:-1])
+    assert sent[-1] == stream_payload
+
+
+@pytest.mark.asyncio
+async def test_send_by_session_splits_long_proactive_text() -> None:
+    adapter = QQOfficialPlatformAdapter(
+        {
+            "id": "qq-official-test",
+            "appid": "123",
+            "secret": "secret",
+            "enable_group_c2c": True,
+            "enable_guild_direct_message": False,
+        },
+        {},
+        asyncio.Queue(),
+    )
+    adapter.client.api = SimpleNamespace(
+        post_group_message=AsyncMock(return_value={"id": "sent-1"}),
+        post_message=AsyncMock(),
+    )
+    adapter._session_scene["group-1"] = "group"
+
+    long_text = "不稀罕。" * 1500
+    await adapter.send_by_session(
+        MessageSession("qq_official", MessageType.GROUP_MESSAGE, "group-1"),
+        MessageChain(chain=[Plain(long_text)]),
+    )
+
+    assert adapter.client.api.post_group_message.await_count > 1
+    sent = [
+        _extract_send_text(kwargs)
+        for _, kwargs in adapter.client.api.post_group_message.await_args_list
+    ]
+    assert all(len(t) <= QQOfficialMessageEvent.QQ_MAX_LENGTH for t in sent)
+    assert "".join(sent) == long_text