Update OpenAI API (#667)

python/sglang/srt/conversation.py

@@ -6,7 +6,7 @@ import dataclasses
 from enum import IntEnum, auto
 from typing import Dict, List, Optional, Tuple, Union
 
-from sglang.srt.openai_protocol import ChatCompletionRequest
+from sglang.srt.openai_api.protocol import ChatCompletionRequest
 
 
 class SeparatorStyle(IntEnum):

python/sglang/srt/openai_api_adapter.py → python/sglang/srt/openai_api/adapter.py

@@ -16,7 +16,7 @@ from sglang.srt.conversation import (
     register_conv_template,
 )
 from sglang.srt.managers.io_struct import GenerateReqInput
-from sglang.srt.openai_protocol import (
+from sglang.srt.openai_api.protocol import (
     ChatCompletionRequest,
     ChatCompletionResponse,
     ChatCompletionResponseChoice,
@@ -106,6 +106,7 @@ async def v1_completions(tokenizer_manager, raw_request: Request):
             "frequency_penalty": request.frequency_penalty,
             "regex": request.regex,
             "n": request.n,
+            "ignore_eos": request.ignore_eos,
         },
         return_logprob=request.logprobs is not None and request.logprobs > 0,
         top_logprobs_num=request.logprobs if request.logprobs is not None else 0,

python/sglang/srt/openai_protocol.py → python/sglang/srt/openai_api/protocol.py

@@ -68,6 +68,7 @@ class CompletionRequest(BaseModel):
 
     # Extra parameters for SRT backend only and will be ignored by OpenAI models.
     regex: Optional[str] = None
+    ignore_eos: Optional[bool] = False
 
 
 class CompletionResponseChoice(BaseModel):

python/sglang/srt/server.py

@@ -39,12 +39,12 @@ from sglang.srt.managers.controller.manager_single import (
 from sglang.srt.managers.detokenizer_manager import start_detokenizer_process
 from sglang.srt.managers.io_struct import GenerateReqInput
 from sglang.srt.managers.tokenizer_manager import TokenizerManager
-from sglang.srt.openai_api_adapter import (
+from sglang.srt.openai_api.adapter import (
     load_chat_template_for_openai_api,
     v1_chat_completions,
     v1_completions,
 )
-from sglang.srt.openai_protocol import ModelCard, ModelList
+from sglang.srt.openai_api.protocol import ModelCard, ModelList
 from sglang.srt.server_args import PortArgs, ServerArgs
 from sglang.srt.utils import (
     API_KEY_HEADER_NAME,

python/sglang/test/test_conversation.py

 from sglang.srt.conversation import generate_chat_conv
-from sglang.srt.managers.openai_protocol import (
+from sglang.srt.managers.openai_api.protocol import (
     ChatCompletionMessageContentImagePart,
     ChatCompletionMessageContentImageURL,
     ChatCompletionMessageContentTextPart,

python/sglang/test/test_openai_protocol.py

-from sglang.srt.managers.openai_protocol import (
+from sglang.srt.managers.openai_api.protocol import (
     ChatCompletionMessageContentImagePart,
     ChatCompletionMessageContentImageURL,
     ChatCompletionMessageContentTextPart,