[OAI] Add rid tracing for v1/embeddings and fix rid type in Chat (#6397)

python/sglang/srt/layers/attention/flashattention_backend.py

@@ -918,8 +918,8 @@ class FlashAttentionBackend(AttentionBackend):
             and local_attn_metadata is not None
             and (hasattr(layer, "use_irope") and layer.use_irope)
         )
-        
-        # When Spec Decode enabled, forward_decode would be called with two mode:

+
+        # When Spec Decode enabled, forward_decode would be called with two mode:
         # 1. DRAFT_DECODE: we enable cascade attention when top_k > 1
         # 2. IDLE: we don’t need cascade attention, spec_info will be none in this case
         use_cascade_attn = forward_batch.spec_info is not None and self.topk > 1

python/sglang/srt/openai_api/adapter.py

@@ -1827,8 +1827,10 @@ def v1_embedding_request(all_requests, tokenizer_manager):
             )
         else:
             prompt_kwargs = {"input_ids": prompts}
+    request_ids = [req.request_id for req in all_requests]
 
     adapted_request = EmbeddingReqInput(
+        rid=request_ids,
         **prompt_kwargs,
     )
 

python/sglang/srt/openai_api/protocol.py

@@ -393,7 +393,7 @@ class ChatCompletionRequest(BaseModel):
     chat_template_kwargs: Optional[Dict] = None
 
     # The request id.
-    rid: Optional[Union[List[str], str]] = None
+    rid: Optional[str] = None
 
     # For PD disaggregation
     bootstrap_host: Optional[str] = None
@@ -469,6 +469,9 @@ class EmbeddingRequest(BaseModel):
     dimensions: int = None
     user: Optional[str] = None
 
+    # The request id.
+    rid: Optional[str] = None
+
 
 class EmbeddingObject(BaseModel):
     embedding: List[float]