wavetermdev/waveterm · error

invalid input format

Error message

invalid input format

What it means

This error comes from the ToolAnyCallback of a sum-style AI tool in tools.go. The callback type-asserts its raw `input any` argument to map[string]any; if the runtime input is not a JSON object (map), the assertion fails and this error is returned to the LLM caller. It guards against malformed tool invocations before any parameter extraction.

Source

Thrown at pkg/aiusechat/tools.go:289

		Strict:      true,
		InputSchema: map[string]any{
			"type": "object",
			"properties": map[string]any{
				"values": map[string]any{
					"type": "array",
					"items": map[string]any{
						"type": "integer",
					},
					"description": "Array of numbers to add together",
				},
			},
			"required":             []string{"values"},
			"additionalProperties": false,
		},
		ToolAnyCallback: func(input any, toolUseData *uctypes.UIMessageDataToolUse) (any, error) {
			inputMap, ok := input.(map[string]any)
			if !ok {
				return nil, fmt.Errorf("invalid input format")
			}

			valuesInterface, ok := inputMap["values"]
			if !ok {
				return nil, fmt.Errorf("missing values parameter")
			}

			valuesSlice, ok := valuesInterface.([]any)
			if !ok {
				return nil, fmt.Errorf("values must be an array")
			}

			if len(valuesSlice) == 0 {
				return 0, nil
			}

			sum := 0
			for i, val := range valuesSlice {

View on GitHub (pinned to a4447c1563)

Solutions

1. Ensure the tool call arguments are a JSON object wrapping the array, e.g. {"values": [1,2,3]} instead of [1,2,3]
2. Check the model/SDK version; newer models with strict tool-calling honor the declared schema's additionalProperties/type constraints better
3. If invoking programmatically, pass a map[string]any (or object that re-marshals to a JSON object), not a slice or scalar

Example fix

// before
anySum([1, 2, 3])
// after
anySum({"values": [1, 2, 3]})

if v, ok := raw.(map[string]any); !ok { return fmt.Errorf("tool input must be a JSON object") }; if _, ok := v["values"]; !ok { return fmt.Errorf("missing values") }; if _, ok := v["values"].([]any); !ok { return fmt.Errorf("values must be an array") }

func asObjectMap(input any) (map[string]any, bool) { m, ok := input.(map[string]any); return m, ok }

result, err := tool.Call(input); if err != nil { if strings.Contains(err.Error(), "invalid input format") { /* re-invoke model with corrected schema example */ } return err }

When it happens

Trigger: The AI/LLM invokes the tool with input that is not a JSON object — e.g. a bare array, string, number, or nil is passed as the tool arguments instead of an object like {"values":[1,2,3]}.

Common situations: Model-generated tool call arguments that are JSON arrays or strings instead of objects; older model versions ignoring the JSON schema's type:object constraint; a caller invoking the tool programmatically passing a non-map value; JSON deserialization layers that hand raw scalars through.

Related errors

• missing values parameter
• values must be an array
• value at index %d is not a number
• invalid input format: %w
• Invalid path part: ${pathPart}

AI-assisted analysis of wavetermdev/waveterm@a4447c1563 (2026-09-01). Data as JSON: /api/errors/b65232e2a2bbcf9a. Report an issue: GitHub.