github/DB-GPT
1
0
Fork 0
mirror of https://github.com/csunny/DB-GPT.git synced 2025-09-05 02:51:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat(core): Support multi round conversation operator (#986)

Browse Source
This commit is contained in:
Fangyin Cheng
2023-12-27 23:26:28 +08:00
committed by GitHub
parent 9aec636b02
commit b13d3f6d92
63 changed files with 2011 additions and 314 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
dbgpt/model/cluster/apiserver/api.py
View File

@@ -16,7 +16,6 @@ from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import StreamingResponse, JSONResponse
from fastapi.security.http import HTTPAuthorizationCredentials, HTTPBearer
from pydantic import BaseSettings
from fastchat.protocol.openai_api_protocol import (
ChatCompletionResponse,
@@ -42,6 +41,7 @@ from dbgpt.model.parameter import ModelAPIServerParameters, WorkerType
from dbgpt.model.cluster import ModelRegistry
from dbgpt.model.cluster.manager_base import WorkerManager, WorkerManagerFactory
from dbgpt.util.utils import setup_logging
from dbgpt._private.pydantic import BaseModel
logger = logging.getLogger(__name__)
@@ -52,7 +52,7 @@ class APIServerException(Exception):
self.message = message
class APISettings(BaseSettings):
class APISettings(BaseModel):
api_keys: Optional[List[str]] = None

13
dbgpt/model/utils/chatgpt_utils.py
View File

@@ -24,6 +24,7 @@ from dbgpt.core.interface.llm import ModelMetadata, LLMClient
from dbgpt.core.interface.llm import ModelOutput, ModelRequest
from dbgpt.model.cluster.client import DefaultLLMClient
from dbgpt.model.cluster import WorkerManagerFactory
from dbgpt._private.pydantic import model_to_json
if TYPE_CHECKING:
import httpx
@@ -175,6 +176,9 @@ class OpenAILLMClient(LLMClient):
async def generate(self, request: ModelRequest) -> ModelOutput:
messages = request.to_openai_messages()
payload = self._build_request(request)
logger.info(
f"Send request to openai, payload: {payload}\n\n messages:\n{messages}"
)
try:
chat_completion = await self.client.chat.completions.create(
messages=messages, **payload
@@ -193,6 +197,9 @@ class OpenAILLMClient(LLMClient):
) -> AsyncIterator[ModelOutput]:
messages = request.to_openai_messages()
payload = self._build_request(request, True)
logger.info(
f"Send request to openai, payload: {payload}\n\n messages:\n{messages}"
)
try:
chat_completion = await self.client.chat.completions.create(
messages=messages, **payload
@@ -321,7 +328,7 @@ async def _to_openai_stream(
chunk = ChatCompletionStreamResponse(
id=id, choices=[choice_data], model=model or ""
)
yield f"data: {chunk.json(exclude_unset=True, ensure_ascii=False)}\n\n"
yield f"data: {model_to_json(chunk, exclude_unset=True, ensure_ascii=False)}\n\n"
previous_text = ""
finish_stream_events = []
@@ -356,7 +363,7 @@ async def _to_openai_stream(
if model_output.finish_reason is not None:
finish_stream_events.append(chunk)
continue
yield f"data: {chunk.json(exclude_unset=True, ensure_ascii=False)}\n\n"
yield f"data: {model_to_json(chunk, exclude_unset=True, ensure_ascii=False)}\n\n"
for finish_chunk in finish_stream_events:
yield f"data: {finish_chunk.json(exclude_none=True, ensure_ascii=False)}\n\n"
yield f"data: {model_to_json(finish_chunk, exclude_none=True, ensure_ascii=False)}\n\n"
yield "data: [DONE]\n\n"