[Frontend] Delegate preprocessing to OpenAIServingRender (#36483)

Signed-off-by: Sage Ahrac <sagiahrak@gmail.com>
This commit is contained in:
Sage
2026-03-13 00:39:43 -07:00
committed by GitHub
parent a4ad9db541
commit a2268617cf
10 changed files with 203 additions and 196 deletions
+26 -21
View File
@@ -72,6 +72,29 @@ async def init_generate_state(
tool_server = None
resolved_chat_template = load_chat_template(args.chat_template)
# Render endpoints are always backed by OpenAIServingRender so that
# /v1/chat/completions/render and /v1/completions/render work on both
# generate-mode and render-only servers.
# It is created first so that OpenAIServingChat and OpenAIServingCompletion
# can delegate their preprocessing logic to it.
from vllm.entrypoints.serve.render.serving import OpenAIServingRender
state.openai_serving_render = OpenAIServingRender(
model_config=engine_client.model_config,
renderer=engine_client.renderer,
io_processor=engine_client.io_processor,
model_registry=state.openai_serving_models.registry,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
trust_request_chat_template=args.trust_request_chat_template,
enable_auto_tools=args.enable_auto_tool_choice,
exclude_tools_when_tool_choice_none=args.exclude_tools_when_tool_choice_none,
tool_parser=args.tool_call_parser,
default_chat_template_kwargs=args.default_chat_template_kwargs,
log_error_stack=args.log_error_stack,
)
state.openai_serving_responses = (
OpenAIServingResponses(
engine_client,
@@ -96,6 +119,7 @@ async def init_generate_state(
engine_client,
state.openai_serving_models,
args.response_role,
openai_serving_render=state.openai_serving_render,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
@@ -120,6 +144,7 @@ async def init_generate_state(
OpenAIServingCompletion(
engine_client,
state.openai_serving_models,
openai_serving_render=state.openai_serving_render,
request_logger=request_logger,
return_tokens_as_token_ids=args.return_tokens_as_token_ids,
enable_prompt_tokens_details=args.enable_prompt_tokens_details,
@@ -133,6 +158,7 @@ async def init_generate_state(
engine_client,
state.openai_serving_models,
args.response_role,
openai_serving_render=state.openai_serving_render,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
@@ -159,24 +185,3 @@ async def init_generate_state(
if "generate" in supported_tasks
else None
)
# Render endpoints are always backed by OpenAIServingRender so that
# /v1/chat/completions/render and /v1/completions/render work on both
# generate-mode and render-only servers.
from vllm.entrypoints.serve.render.serving import OpenAIServingRender
state.openai_serving_render = OpenAIServingRender(
model_config=engine_client.model_config,
renderer=engine_client.renderer,
io_processor=engine_client.io_processor,
model_registry=state.openai_serving_models.registry,
request_logger=request_logger,
chat_template=resolved_chat_template,
chat_template_content_format=args.chat_template_content_format,
trust_request_chat_template=args.trust_request_chat_template,
enable_auto_tools=args.enable_auto_tool_choice,
exclude_tools_when_tool_choice_none=args.exclude_tools_when_tool_choice_none,
tool_parser=args.tool_call_parser,
default_chat_template_kwargs=args.default_chat_template_kwargs,
log_error_stack=args.log_error_stack,
)