How to return a vector []float32 from a tool call

[Dabnis]

Pre-submission Checklist

• I have checked that this question would not be more appropriate as an issue in a specific repository
• I have searched existing discussions and documentation for answers

Question Category

• Protocol Specification
• SDK Usage
• Server Implementation
• General Implementation
• Documentation
• Other

Your Question

I am trying to return a []float32 vector from a tool call, this is my code:

	buf := new(bytes.Buffer)
	if err := binary.Write(buf, binary.LittleEndian, v); err != nil {
		slog.ErrorContext(job.ctx, "Failed to convert embedding to bytes", "error", err)
		job.error <- fmt.Errorf("internal error converting embedding: %w", err)
		return
	}
	embeddingBytes := buf.Bytes()

	encodedData := base64.StdEncoding.EncodeToString(embeddingBytes)
	dataURI := "data:application/octet-stream;base64," + encodedData

	// Send the result back to the dispatcher as a CallToolResult.
	job.result <- &mcp.CallToolResultFor[any]{
		Content: []mcp.Content{
			&mcp.EmbeddedResource{
				Resource: &mcp.ResourceContents{
					URI: dataURI,
				},
			},
		},
	}

which fails with many errors!
Any suggestions appreciated.

[Dabnis]

I'm not 100% that this is the correct method, but the following works!

embeddingJSON, err := json.Marshal(v)
	if err != nil {
		slog.ErrorContext(job.ctx, "Failed to marshal embedding response", "error", err)
		job.error <- fmt.Errorf("internal error marshalling embedding: %w", err)
		return
	}

	// Send the result back to the dispatcher as a CallToolResult.
	job.result <- &mcp.CallToolResultFor[any]{
		Content: []mcp.Content{
			&mcp.TextContent{
				Text: string(embeddingJSON),
			},
		},
	}

[jba]

I would use the StructuredContent field to return it as JSON.
In our current API, your tool handler would look like:

func(context.Context, *ServerRequest[*CallToolParamsFor[In]]) (*CallToolResultFor[[]float64], error)

where In is whatever the input to your function is.

But note, this API will be changing in v0.3.0, due out Friday.

[Dabnis]

Thanks for the feedback, appreciated as always :)