add support to enable lora with embedding models (#17780)

Co-authored-by: Vedant Jhaveri <vjhaveri@linkedin.com>

python/sglang/srt/entrypoints/openai/serving_embedding.py

@@ -126,12 +126,24 @@ class OpenAIServingEmbedding(OpenAIServingBase):
             # Other types (should not happen but handle gracefully)
             prompt_kwargs = {"input_ids": prompt}
 
+        # Resolve LoRA adapter from model parameter or explicit lora_path
+        lora_path = self._resolve_lora_path(request.model, request.lora_path)
+        if lora_path:
+            first_adapter = (
+                lora_path
+                if isinstance(lora_path, str)
+                else next((a for a in lora_path if a), None)
+            )
+            if first_adapter:
+                self._validate_lora_enabled(first_adapter)
+
         adapted_request = EmbeddingReqInput(
             **prompt_kwargs,
             rid=request.rid,
             priority=request.priority,
             routing_key=self.extract_routing_key(raw_request),
             dimensions=request.dimensions,
+            lora_path=lora_path,
         )
 
         return adapted_request, request