[Frontend] Introduce Renderer for processing chat messages (using ModelConfig) (#30200)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-01-22 12:44:22 +00:00
committed by GitHub
parent 421012b63a
commit d117a4d1a9
48 changed files with 2141 additions and 1585 deletions
+2 -4
View File
@@ -78,13 +78,10 @@ class EmbeddingMixin(OpenAIServing):
try:
ctx.lora_request = self._maybe_get_adapters(ctx.request)
tokenizer = await self.engine_client.get_tokenizer()
renderer = self._get_renderer(tokenizer)
if isinstance(ctx.request, EmbeddingChatRequest):
_, ctx.engine_prompts = await self._preprocess_chat(
ctx.request,
tokenizer,
self.renderer,
ctx.request.messages,
chat_template=ctx.request.chat_template or ctx.chat_template,
chat_template_content_format=ctx.chat_template_content_format,
@@ -93,6 +90,7 @@ class EmbeddingMixin(OpenAIServing):
add_special_tokens=ctx.request.add_special_tokens,
)
else:
renderer = self._get_completion_renderer()
ctx.engine_prompts = await renderer.render_prompt(
prompt_or_prompts=ctx.request.input,
config=self._build_render_config(ctx.request),