Spaces:
Sleeping
Sleeping
File size: 9,458 Bytes
99bf030 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 |
import inspect
import json
import inspect
import json
import logging
import time
import uuid
from typing import Generator, Iterator, AsyncGenerator, Optional
from aworld.core.task import Task
from aworld.utils.common import get_local_ip
from fastapi import status, HTTPException
from fastapi.concurrency import run_in_threadpool
from pydantic import BaseModel
from starlette.responses import StreamingResponse
from aworldspace.base import AGENT_SPACE
from aworldspace.utils.utils import get_last_user_message
from base import OpenAIChatCompletionForm
async def generate_openai_chat_completion(form_data: OpenAIChatCompletionForm):
messages = [message.model_dump() for message in form_data.messages]
user_message = get_last_user_message(messages)
PIPELINES = await AGENT_SPACE.get_agents_meta()
PIPELINE_MODULES = await AGENT_SPACE.get_agent_modules()
if (
form_data.model not in PIPELINES
or PIPELINES[form_data.model]["type"] == "filter"
):
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,
detail=f"Pipeline {form_data.model} not found",
)
def job():
pipeline = PIPELINES[form_data.model]
pipeline_id = form_data.model
if pipeline["type"] == "manifold":
manifold_id, pipeline_id = pipeline_id.split(".", 1)
pipe = PIPELINE_MODULES[manifold_id].pipe
else:
pipe = PIPELINE_MODULES[pipeline_id].pipe
def process_line(model, line):
if isinstance(line, Task):
task_output_meta = line.outputs._metadata
line = openai_chat_chunk_message_template(model, "", task_output_meta=task_output_meta)
return f"data: {json.dumps(line)}\n\n"
if isinstance(line, BaseModel):
line = line.model_dump_json()
line = f"data: {line}"
if isinstance(line, dict):
line = f"data: {json.dumps(line)}"
try:
line = line.decode("utf-8")
except Exception:
pass
if line.startswith("data:"):
return f"{line}\n\n"
else:
line = openai_chat_chunk_message_template(model, line)
return f"data: {json.dumps(line)}\n\n"
if form_data.stream:
async def stream_content():
async def execute_pipe(_pipe):
if inspect.iscoroutinefunction(_pipe):
return await _pipe(user_message=user_message,
model_id=pipeline_id,
messages=messages,
body=form_data.model_dump())
else:
return _pipe(user_message=user_message,
model_id=pipeline_id,
messages=messages,
body=form_data.model_dump())
try:
res = await execute_pipe(pipe)
# Directly return if the response is a StreamingResponse
if isinstance(res, StreamingResponse):
async for data in res.body_iterator:
yield data
return
if isinstance(res, dict):
yield f"data: {json.dumps(res)}\n\n"
return
except Exception as e:
logging.error(f"Error: {e}")
import traceback
traceback.print_exc()
yield f"data: {json.dumps({'error': {'detail': str(e)}})}\n\n"
return
if isinstance(res, str):
message = openai_chat_chunk_message_template(form_data.model, res)
yield f"data: {json.dumps(message)}\n\n"
if isinstance(res, Iterator):
for line in res:
yield process_line(form_data.model, line)
if isinstance(res, AsyncGenerator):
async for line in res:
yield process_line(form_data.model, line)
logging.info(f"AsyncGenerator end...")
if isinstance(res, str) or isinstance(res, Generator) or isinstance(res, AsyncGenerator):
finish_message = openai_chat_chunk_message_template(
form_data.model, ""
)
finish_message["choices"][0]["finish_reason"] = "stop"
print(f"Pipe-Dataline:::: DONE")
yield f"data: {json.dumps(finish_message)}\n\n"
yield "data: [DONE]"
return StreamingResponse(stream_content(), media_type="text/event-stream")
else:
res = pipe(
user_message=user_message,
model_id=pipeline_id,
messages=messages,
body=form_data.model_dump(),
)
logging.info(f"stream:false:{res}")
if isinstance(res, dict):
return res
elif isinstance(res, BaseModel):
return res.model_dump()
else:
message = ""
if isinstance(res, str):
message = res
if isinstance(res, Generator):
for stream in res:
message = f"{message}{stream}"
logging.info(f"stream:false:{message}")
return {
"id": f"{form_data.model}-{str(uuid.uuid4())}",
"object": "chat.completion",
"created": int(time.time()),
"model": form_data.model,
"choices": [
{
"index": 0,
"message": {
"role": "assistant",
"content": message,
},
"logprobs": None,
"finish_reason": "stop",
}
],
}
return await run_in_threadpool(job)
async def call_pipeline(form_data: OpenAIChatCompletionForm):
messages = [message.model_dump() for message in form_data.messages]
user_message = get_last_user_message(messages)
PIPELINES = await AGENT_SPACE.get_agents_meta()
PIPELINE_MODULES = await AGENT_SPACE.get_agent_modules()
if (
form_data.model not in PIPELINES
or PIPELINES[form_data.model]["type"] == "filter"
):
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,
detail=f"Pipeline {form_data.model} not found",
)
pipeline = PIPELINES[form_data.model]
pipeline_id = form_data.model
if pipeline["type"] == "manifold":
manifold_id, pipeline_id = pipeline_id.split(".", 1)
pipe = PIPELINE_MODULES[manifold_id].pipe
else:
pipe = PIPELINE_MODULES[pipeline_id].pipe
if form_data.stream:
async def execute_pipe(_pipe):
if inspect.iscoroutinefunction(_pipe):
return await _pipe(user_message=user_message,
model_id=pipeline_id,
messages=messages,
body=form_data.model_dump())
else:
return _pipe(user_message=user_message,
model_id=pipeline_id,
messages=messages,
body=form_data.model_dump())
res = await execute_pipe(pipe)
return res
else:
if not inspect.iscoroutinefunction(pipe):
return await run_in_threadpool(
pipe,
user_message=user_message,
model_id=pipeline_id,
messages=messages,
body=form_data.model_dump()
)
else:
return await pipe(
user_message=user_message,
model_id=pipeline_id,
messages=messages,
body=form_data.model_dump()
)
def openai_chat_chunk_message_template(
model: str,
content: Optional[str] = None,
tool_calls: Optional[list[dict]] = None,
usage: Optional[dict] = None,
**kwargs
) -> dict:
template = openai_chat_message_template(model, **kwargs)
template["object"] = "chat.completion.chunk"
template["choices"][0]["index"] = 0
template["choices"][0]["delta"] = {}
if content:
template["choices"][0]["delta"]["content"] = content
if tool_calls:
template["choices"][0]["delta"]["tool_calls"] = tool_calls
if not content and not tool_calls:
template["choices"][0]["finish_reason"] = "stop"
if usage:
template["usage"] = usage
return template
def openai_chat_message_template(model: str, **kwargs):
return {
"id": f"{model}-{str(uuid.uuid4())}",
"created": int(time.time()),
"model": model,
"node_id": get_local_ip(),
"task_output_meta": kwargs.get("task_output_meta"),
"choices": [{"index": 0, "logprobs": None, "finish_reason": None}],
} |