PoC: logprobs support for vLLM backend

[rmccorm4]

Overview:

Draft to test out the current state of logprobs support

• Frontend already supports logprobs
• Backend is not propagating out the logprobs information in the LLMEngineOutput responses that get yielded

Details:

Example server setup:

uv venv venv
source venv/bin/activate

uv pip install ai-dynamo[vllm]==0.7.0

python -m dynamo.frontend &

python -m dynamo.vllm --model Qwen/Qwen3-0.6B --enforce-eager --connector none &

Example completions request:

MODEL="Qwen/Qwen3-0.6B"

curl -X POST http://localhost:8000/v1/completions \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "'${MODEL}'",
    "prompt": "What is the test plan?",
    "logprobs": 2,
    "temperature": 0,
    "max_tokens": 10
  }' | jq

Example chat completions request:

MODEL="Qwen/Qwen3-0.6B"

curl -X POST http://localhost:8000/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "'${MODEL}'",
    "messages": [{"role": "user", "content": "What is the test plan?"}],
    "logprobs": true,
    "top_logprobs": 2,
    "temperature": 0,
    "max_tokens": 50
  }' | jq

Where should the reviewer start?

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

• closes GitHub issue: #xxx