MCP DeepInfra AI Tools Server

Instructions

Implementation Reference

• src/mcp_deepinfra/server.py:154-182 (handler)

  The handler function for the 'object_detection' tool. It takes an image URL, prompts a vision-language model (gpt-4o-mini by default) to detect objects, and returns the model's response as a string, formatted as JSON.

  async def object_detection(image_url: str) -> str:
      """Detect objects in an image using DeepInfra OpenAI-compatible API with multimodal model."""
      model = DEFAULT_MODELS["object_detection"]
      try:
          response = await client.chat.completions.create(
              model=model,
              messages=[
                  {
                      "role": "user",
                      "content": [
                          {
                              "type": "text",
                              "text": "Analyze this image and detect all objects present. Provide a detailed list of objects you can see, their approximate locations if possible, and confidence scores. Format as JSON."
                          },
                          {
                              "type": "image_url",
                              "image_url": {"url": image_url}
                          }
                      ]
                  }
              ],
              max_tokens=500,
          )
          if response.choices:
              return response.choices[0].message.content
          else:
              return "No objects detected"
      except Exception as e:
          return f"Error detecting objects: {type(e).__name__}: {str(e)}"

• src/mcp_deepinfra/server.py:152-153 (registration)

  Registers the object_detection tool with the FastMCP server using the @app.tool() decorator, conditionally based on the ENABLED_TOOLS configuration.

  if "all" in ENABLED_TOOLS or "object_detection" in ENABLED_TOOLS:
      @app.tool()

• src/mcp_deepinfra/server.py:37-37 (helper)

  Configuration for the default model used by the object_detection tool.

  "object_detection": os.getenv("MODEL_OBJECT_DETECTION", "openai/gpt-4o-mini"),

• src/mcp_deepinfra/server.py:154-155 (schema)

  Function signature and docstring defining the input (image_url: str) and output (str) schema for the tool.

  async def object_detection(image_url: str) -> str:
      """Detect objects in an image using DeepInfra OpenAI-compatible API with multimodal model."""