add ChatTemplateKwargs to ChatCompletionRequest

Author: JustaCai

chat.go

@@ -275,6 +275,11 @@ type ChatCompletionRequest struct {
 	Metadata map[string]string `json:"metadata,omitempty"`
 	// Configuration for a predicted output.
 	Prediction *Prediction `json:"prediction,omitempty"`
+	// ChatTemplateKwargs provides a way to add non-standard parameters to the request body.
+	// https://github.com/vllm-project/vllm/blob/05a4324f8e3932c25554791ff248e3e0200eef92/vllm/entrypoints/openai/protocol.py#L320
+	// Additional kwargs to pass to the template renderer. Will be accessible by the chat template.
+	// Such as think mode for qwen3. https://qwen.readthedocs.io/en/latest/deployment/vllm.html#thinking-non-thinking-modes
+	ChatTemplateKwargs map[string]any `json:"chat_template_kwargs,omitempty"`
 }
 
 type StreamOptions struct {