Currently, the simulator does not support multimodal request payloads, as used by frameworks like Llama Stack or any OpenAI-compatible client that sends structured content blocks (e.g., text, image_url, video_url).

These clients send a content array of objects instead of a plain string. When such a request is received, the simulator fails to parse the body, expecting content to be of type string, not []object.

🔁 Example Request (Valid OpenAI format)

curl -k -X POST http://localhost:8000/v1/chat/completions \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "inference-simulator-maas",
    "messages": [
      {
        "role": "user",
        "content": [
          { "type": "text", "text": "Can you describe this image?" },
          { "type": "image_url", "url": "https://example.com/image.png" }
        ]
      }
    ]
  }'

❌ Current Result

E0604 09:16:14.284161       1 simulator.go:264] "failed to read and parse request body" err="json: cannot unmarshal array into Go struct field message.messages.content of type string"

✅ Expected Behavior

The simulator should:

• Accept content as an array of typed blocks (type: "text", type: "image_url", etc.)
• Gracefully handle or skip unsupported types (e.g., images or videos) if not yet implemented
• Provide fallback or mock behavior where needed for inference simulation