Function tooling like in vllm

Hi there!
Is there a possibility to use this feature like in vllm https://docs.vllm.ai/en/stable/features/tool_calling.html
in TensorRT-LLM?

vllm serve deepseek--enable-auto-tool-choice  --tool-call-parser deepseek_v3 --chat-template tool_chat_template_deepseekr1.jinja 

https://github.com/vllm-project/vllm/blob/main/examples/tool_chat_template_deepseekv3.jinja

I did not find it in docs.
We use nvidia/DeepSeek-R1-0528-FP4 and nvidia/DeepSeek-V3-0324-FP4

trtllm-serve
  --backend pytorch \
  --tp_size 8 \
  --ep_size 2 \
  --max_num_tokens 32768 \
  --trust_remote_code \
  --extra_llm_api_options /root/extra-llm-api-config.yml \
  --kv_cache_free_gpu_memory_fraction 0.85'

extra-llm-api-config.yml

cuda_graph_config: {}
moe_backend: TRTLLM
speculative_config:
    decoding_type: MTP
    num_nextn_predict_layers: 3

request
google adk->litellm->trtllm-serve