mirror of
https://github.com/RYDE-WORK/Langchain-Chatchat.git
synced 2026-01-19 13:23:16 +08:00
9ff7bef2c2
* 新功能： - 知识库管理界面支持查看、编辑、删除向量库文档。暂不支持增加（aggrid添加新行比较麻烦，需要另外实现） - 去除知识库管理界面中重建知识库和删除知识库按钮，建议用户到终端命令操作 修复： - 所有与知识库名称、文件名称有关的数据库操作函数都改成大小写不敏感，所有路径统一为 posix 风格，避免因路径文本不一致导致数据重复和操作失效 （close #2232） 开发者： - 添加 update_docs_by_id 函数与 API 接口。当前仅支持 FAISS，暂时未用到，未将来对知识库做更细致的修改做准备 - 统一 DocumentWithScore 与 DocumentWithVsId - FAISS 返回的 Document.metadata 中包含 ID， 方便后续查找比对 - /knowledge_base/search_docs 接口支持 file_name, metadata 参数，可以据此检索文档 * fix bug
104 lines
5.3 KiB
Python
104 lines
5.3 KiB
Python
from fastapi import Body
|
||
from sse_starlette.sse import EventSourceResponse
|
||
from configs import LLM_MODELS, TEMPERATURE
|
||
from server.utils import wrap_done, get_ChatOpenAI
|
||
from langchain.chains import LLMChain
|
||
from langchain.callbacks import AsyncIteratorCallbackHandler
|
||
from typing import AsyncIterable
|
||
import asyncio
|
||
import json
|
||
from langchain.prompts.chat import ChatPromptTemplate
|
||
from typing import List, Optional, Union
|
||
from server.chat.utils import History
|
||
from langchain.prompts import PromptTemplate
|
||
from server.utils import get_prompt_template
|
||
from server.memory.conversation_db_buffer_memory import ConversationBufferDBMemory
|
||
from server.db.repository import add_message_to_db
|
||
from server.callback_handler.conversation_callback_handler import ConversationCallbackHandler
|
||
|
||
|
||
async def chat(query: str = Body(..., description="用户输入", examples=["恼羞成怒"]),
|
||
conversation_id: str = Body("", description="对话框ID"),
|
||
history_len: int = Body(-1, description="从数据库中取历史消息的数量"),
|
||
history: Union[int, List[History]] = Body([],
|
||
description="历史对话，设为一个整数可以从数据库中读取历史消息",
|
||
examples=[[
|
||
{"role": "user",
|
||
"content": "我们来玩成语接龙，我先来，生龙活虎"},
|
||
{"role": "assistant", "content": "虎头虎脑"}]]
|
||
),
|
||
stream: bool = Body(False, description="流式输出"),
|
||
model_name: str = Body(LLM_MODELS[0], description="LLM 模型名称。"),
|
||
temperature: float = Body(TEMPERATURE, description="LLM 采样温度", ge=0.0, le=2.0),
|
||
max_tokens: Optional[int] = Body(None, description="限制LLM生成Token数量，默认None代表模型最大值"),
|
||
# top_p: float = Body(TOP_P, description="LLM 核采样。勿与temperature同时设置", gt=0.0, lt=1.0),
|
||
prompt_name: str = Body("default", description="使用的prompt模板名称(在configs/prompt_config.py中配置)"),
|
||
):
|
||
async def chat_iterator() -> AsyncIterable[str]:
|
||
nonlocal history, max_tokens
|
||
callback = AsyncIteratorCallbackHandler()
|
||
callbacks = [callback]
|
||
memory = None
|
||
|
||
# 负责保存llm response到message db
|
||
message_id = add_message_to_db(chat_type="llm_chat", query=query, conversation_id=conversation_id)
|
||
conversation_callback = ConversationCallbackHandler(conversation_id=conversation_id, message_id=message_id,
|
||
chat_type="llm_chat",
|
||
query=query)
|
||
callbacks.append(conversation_callback)
|
||
|
||
if isinstance(max_tokens, int) and max_tokens <= 0:
|
||
max_tokens = None
|
||
|
||
model = get_ChatOpenAI(
|
||
model_name=model_name,
|
||
temperature=temperature,
|
||
max_tokens=max_tokens,
|
||
callbacks=callbacks,
|
||
)
|
||
|
||
if history: # 优先使用前端传入的历史消息
|
||
history = [History.from_data(h) for h in history]
|
||
prompt_template = get_prompt_template("llm_chat", prompt_name)
|
||
input_msg = History(role="user", content=prompt_template).to_msg_template(False)
|
||
chat_prompt = ChatPromptTemplate.from_messages(
|
||
[i.to_msg_template() for i in history] + [input_msg])
|
||
elif conversation_id and history_len > 0: # 前端要求从数据库取历史消息
|
||
# 使用memory 时必须 prompt 必须含有memory.memory_key 对应的变量
|
||
prompt = get_prompt_template("llm_chat", "with_history")
|
||
chat_prompt = PromptTemplate.from_template(prompt)
|
||
# 根据conversation_id 获取message 列表进而拼凑 memory
|
||
memory = ConversationBufferDBMemory(conversation_id=conversation_id,
|
||
llm=model,
|
||
message_limit=history_len)
|
||
else:
|
||
prompt_template = get_prompt_template("llm_chat", prompt_name)
|
||
input_msg = History(role="user", content=prompt_template).to_msg_template(False)
|
||
chat_prompt = ChatPromptTemplate.from_messages([input_msg])
|
||
|
||
chain = LLMChain(prompt=chat_prompt, llm=model, memory=memory)
|
||
|
||
# Begin a task that runs in the background.
|
||
task = asyncio.create_task(wrap_done(
|
||
chain.acall({"input": query}),
|
||
callback.done),
|
||
)
|
||
|
||
if stream:
|
||
async for token in callback.aiter():
|
||
# Use server-sent-events to stream the response
|
||
yield json.dumps(
|
||
{"text": token, "message_id": message_id},
|
||
ensure_ascii=False)
|
||
else:
|
||
answer = ""
|
||
async for token in callback.aiter():
|
||
answer += token
|
||
yield json.dumps(
|
||
{"text": answer, "message_id": message_id},
|
||
ensure_ascii=False)
|
||
|
||
await task
|
||
|
||
return EventSourceResponse(chat_iterator())
|