Kris
f551745ec2
1. 删除废弃的 timeline_adapter.py 文件 2. 新增历史消息拉取器基类实现 WeChatHistoryFetcher 3. 新增流式消息分片发送和打字指示器支持逻辑
129 lines
3.9 KiB
Python
129 lines
3.9 KiB
Python
from __future__ import annotations
|
||
|
||
import re
|
||
from typing import TYPE_CHECKING
|
||
|
||
from yuxi.channels.models import DeliveryResult
|
||
from yuxi.utils.logging_config import logger
|
||
|
||
if TYPE_CHECKING:
|
||
from .adapter import WeChatAdapter
|
||
|
||
_PARAGRAPH_SPLITTER = re.compile(r"\n{2,}")
|
||
_SENTENCE_BOUNDARY = re.compile(r"[。！？.!?\n]")
|
||
|
||
|
||
class ParagraphChunker:
|
||
def __init__(self, chunk_size: int = 2000, min_chunk_size: int = 100):
|
||
self.chunk_size = chunk_size
|
||
self.min_chunk_size = min_chunk_size
|
||
self._buffer: list[str] = []
|
||
|
||
def feed(self, text: str) -> list[str]:
|
||
self._buffer.append(text)
|
||
accumulated = "".join(self._buffer)
|
||
results: list[str] = []
|
||
|
||
if "\n\n" in accumulated:
|
||
paragraphs = _PARAGRAPH_SPLITTER.split(accumulated)
|
||
if len(paragraphs) > 1:
|
||
for para in paragraphs[:-1]:
|
||
stripped = para.strip()
|
||
if len(stripped) >= self.min_chunk_size:
|
||
results.append(stripped)
|
||
self._buffer = [paragraphs[-1]]
|
||
|
||
return results
|
||
|
||
def flush(self) -> list[str]:
|
||
if not self._buffer:
|
||
return []
|
||
accumulated = "".join(self._buffer)
|
||
self._buffer = []
|
||
if not accumulated.strip():
|
||
return []
|
||
|
||
if len(accumulated) <= self.chunk_size:
|
||
return [accumulated.strip()]
|
||
|
||
results = []
|
||
remaining = accumulated
|
||
while len(remaining) > self.chunk_size:
|
||
split_at = self._find_split_point(remaining, self.chunk_size)
|
||
results.append(remaining[:split_at].strip())
|
||
remaining = remaining[split_at:].strip()
|
||
if remaining.strip():
|
||
results.append(remaining.strip())
|
||
return results
|
||
|
||
def _find_split_point(self, text: str, limit: int) -> int:
|
||
text_slice = text[:limit]
|
||
for sep in ("\n\n", "\n", "。", "！", "？", ". ", "! ", "? "):
|
||
idx = text_slice.rfind(sep)
|
||
if idx > limit // 2:
|
||
return idx + len(sep)
|
||
return limit
|
||
|
||
|
||
class ProgressPrefixFormatter:
|
||
def __init__(self, template: str = "({current}/{total})", enabled: bool = True):
|
||
self._template = template
|
||
self.enabled = enabled
|
||
|
||
def format(self, current: int, total: int) -> str:
|
||
if not self.enabled:
|
||
return ""
|
||
return self._template.format(current=current, total=total)
|
||
|
||
|
||
async def send_stream_chunked(
|
||
adapter: WeChatAdapter,
|
||
chat_id: str,
|
||
text: str,
|
||
finished: bool = False,
|
||
) -> DeliveryResult:
|
||
from yuxi.channels.models import ChannelIdentity, ChannelResponse
|
||
|
||
identity = ChannelIdentity(
|
||
channel_id=adapter.channel_id,
|
||
channel_type=adapter.channel_type,
|
||
channel_chat_id=chat_id,
|
||
channel_user_id="",
|
||
)
|
||
|
||
response = ChannelResponse(identity=identity, content=text)
|
||
try:
|
||
return await adapter.send(response)
|
||
except Exception as e:
|
||
logger.error(f"[WeChat] Chunked send failed for {chat_id}: {e}")
|
||
return DeliveryResult(success=False, error=str(e))
|
||
|
||
|
||
async def send_with_typing(
|
||
adapter: WeChatAdapter,
|
||
chat_id: str,
|
||
text: str,
|
||
typing_enabled: bool = True,
|
||
is_wecom: bool = False,
|
||
) -> DeliveryResult:
|
||
if typing_enabled and is_wecom:
|
||
try:
|
||
await adapter.heartbeat_adapter.send_typing_indicator(adapter.send, chat_id)
|
||
except Exception:
|
||
pass
|
||
|
||
from yuxi.channels.models import ChannelIdentity, ChannelResponse
|
||
|
||
identity = ChannelIdentity(
|
||
channel_id=adapter.channel_id,
|
||
channel_type=adapter.channel_type,
|
||
channel_chat_id=chat_id,
|
||
channel_user_id="",
|
||
)
|
||
|
||
response = ChannelResponse(identity=identity, content=text)
|
||
try:
|
||
return await adapter.send(response)
|
||
except Exception as e:
|
||
return DeliveryResult(success=False, error=str(e))
|