177 lines
6.1 KiB
Python
177 lines
6.1 KiB
Python
import json
|
||
import redis
|
||
import uuid
|
||
import requests
|
||
from fastapi.responses import StreamingResponse
|
||
from fastapi import FastAPI, Query, Request
|
||
from pydantic import BaseModel
|
||
from typing import Optional, List
|
||
|
||
app = FastAPI()
|
||
# 连接到Redis服务器（默认端口是6379）
|
||
redis_instance = redis.Redis(host='localhost', port=6379, db=0)
|
||
mock_time_created = 20230512084843
|
||
mock_id1 = "131424542124"
|
||
mock_id2 = "1231414141541252525"
|
||
mock_user = "abc-123"
|
||
dify_url = "http://<your dify ip>/v1/chat-messages"
|
||
dify_api_key = 'Bearer <your dify api key>'
|
||
|
||
|
||
class Message(BaseModel):
|
||
content: str
|
||
|
||
|
||
class Resquest_Body(BaseModel):
|
||
messages: List[Message] = []
|
||
app_id: str
|
||
user: str
|
||
session_id: str
|
||
is_stream: bool
|
||
|
||
|
||
# 对接dify，非流式传输
|
||
def request_dify_api(question, request, conversation_id):
|
||
# 目标URL
|
||
url = dify_url
|
||
headers_json = {
|
||
'Content-Type': 'application/json',
|
||
'Authorization': dify_api_key
|
||
}
|
||
# 请求的数据
|
||
data = {
|
||
"inputs": {},
|
||
"query": question,
|
||
"response_mode": "blocking",
|
||
"user": mock_user
|
||
}
|
||
|
||
# 发送POST请求
|
||
response = requests.post(url, data=json.dumps(data), headers=headers_json)
|
||
data = json.loads(response)
|
||
return data["answer"]
|
||
|
||
|
||
# 对接dify，流式传输
|
||
async def generate_streaming_messages(question, request, session_id, conversation_id):
|
||
url = dify_url
|
||
headers_json = {
|
||
'Content-Type': 'application/json',
|
||
'Authorization': dify_api_key
|
||
}
|
||
|
||
# 请求的数据
|
||
request_data = {
|
||
"inputs": {},
|
||
"query": question,
|
||
"conversation_id": conversation_id,
|
||
"response_mode": "streaming",
|
||
"user": mock_user
|
||
}
|
||
r = requests.post(url, json=request_data, headers=headers_json, verify=False, stream=True)
|
||
if r.status_code != 200:
|
||
raise Exception(r.text)
|
||
|
||
data = ""
|
||
|
||
# 在这里，我们使用了 iter_lines 方法来逐行读取响应体
|
||
for line in r.iter_lines():
|
||
if await request.is_disconnected():
|
||
# 客户端已主动断开，中止流式输出，避免不必要的问题，打印报错信息方便定位
|
||
print("# 客户端已主动断开，中止流式输出")
|
||
break
|
||
|
||
# 将每一行解码为字符串
|
||
line = line.decode('utf-8').strip()
|
||
|
||
# 接下来我们要处理三种不同的情况：
|
||
# 1. 如果当前行是空行，则表明前一个数据块已接收完毕（即前文提到的，通过两个换行符结束数据块传输），我们可以对该数据块进行反序列化，并打印出对应的 content 内容；
|
||
# 2. 如果当前行为非空行，且以 data: 开头，则表明这是一个数据块传输的开始，我们去除 data: 前缀后，首先判断是否是结束符 [DONE]，如果不是，将数据内容保存到 data 变量；
|
||
# 3. 如果当前行为非空行，但不以 data: 开头，则表明当前行仍然归属上一个正在传输的数据块，我们将当前行的内容追加到 data 变量尾部；
|
||
|
||
if len(line) == 0:
|
||
if data:
|
||
try:
|
||
chunk = json.loads(data)
|
||
except:
|
||
continue
|
||
|
||
if chunk["event"] != "message":
|
||
continue
|
||
# 如果redis里未缓存该session_id，则使用reids关联session_id和conversation_id，保证后续实现多轮对话
|
||
if not redis_instance.exists(session_id):
|
||
conversiation_id = str(chunk["conversation_id"])
|
||
redis_instance.set(session_id, conversiation_id)
|
||
|
||
# 设置redis中key（session_id）的超时时间，一天换算86400s
|
||
redis_instance.expire(session_id, 86400)
|
||
content = chunk["answer"]
|
||
new_data = {
|
||
"id": mock_id2,
|
||
"created": mock_time_created,
|
||
"choices": [{
|
||
"message": {
|
||
"content": content
|
||
}
|
||
}]
|
||
}
|
||
yield f"data: {json.dumps(new_data, ensure_ascii=False)}\n\n"
|
||
|
||
data = "" # 重置 data
|
||
elif line.startswith("data: "):
|
||
data = line.lstrip("data: ")
|
||
|
||
# 当数据块内容为 [DONE] 时，则表明所有数据块已发送完毕，可断开网络连接
|
||
if data == "[DONE]":
|
||
break
|
||
else:
|
||
# 仍然在追加内容时，为其添加一个换行符
|
||
data = data + "\n" + line
|
||
|
||
# 发送结束标记
|
||
yield "data: [DONE]\n\n"
|
||
|
||
|
||
@app.post("/digital-human")
|
||
async def digital_human(body: Resquest_Body, request: Request):
|
||
question = body.messages[0].content
|
||
session_id = body.session_id
|
||
conversation_id = ""
|
||
|
||
# 如果redis_instance里有缓存session_id，则说明为多轮对话，获取对应的dify中的多轮对话关键参数conversation_id
|
||
if redis_instance.exists(session_id):
|
||
# decode()用于将bytes转换为str，因为redis_instance存储的是bytes类型的数据。
|
||
conversation_id = redis_instance.get(session_id).decode("utf-8")
|
||
|
||
# 选择流式传输
|
||
if body.is_stream:
|
||
return StreamingResponse(
|
||
generate_streaming_messages(question, request, session_id, conversation_id),
|
||
media_type="text/event-stream",
|
||
headers={
|
||
"Cache-Control": "no-cache",
|
||
"Connection": "keep-alive",
|
||
}
|
||
)
|
||
|
||
# 选择非流式传输
|
||
answer = request_dify_api(question, request, conversation_id)
|
||
choices = [{
|
||
"index": 0,
|
||
"message": {
|
||
"content": answer
|
||
}
|
||
}]
|
||
return {
|
||
"id": mock_id1,
|
||
"created": mock_time_created,
|
||
"choices": choices
|
||
}
|
||
|
||
|
||
# 启动应用
|
||
if __name__ == "__main__":
|
||
import uvicorn
|
||
|
||
uvicorn.run(app, host="0.0.0.0", port=8000)
|