langchain-ai/deepagents · error · ValueError

task() field `responseSchema` must be an object when provide

Error message

task() field `responseSchema` must be an object when provided

What it means

`_validate_task_payload` rejects a `responseSchema` field that is present but is not an object (Python dict). `responseSchema` is optional and describes the structured output shape the subagent should return. Raised as a ValueError before subagent dispatch.

Source

Thrown at libs/partners/quickjs/langchain_quickjs/_repl.py:566

            raise ValueError(msg)

        subagent_type = payload.get("subagentType")
        if not isinstance(subagent_type, str) or not subagent_type:
            msg = "task() requires non-empty string field `subagentType`"
            raise ValueError(msg)

        raw_label = payload.get("label")
        if raw_label is not None and not isinstance(raw_label, str):
            msg = "task() field `label` must be a string when provided"
            raise ValueError(msg)
        label = raw_label.strip() if isinstance(raw_label, str) else None
        if label == "":
            label = None

        response_schema = payload.get("responseSchema")
        if response_schema is not None and not isinstance(response_schema, dict):
            msg = "task() field `responseSchema` must be an object when provided"
            raise ValueError(msg)

        return description, subagent_type, label, response_schema

    async def _ainvoke_task_on_outer_loop(
        self,
        payload: dict[str, Any],
        *,
        state: _PTCState,
    ) -> Any:
        """Validate JS `task()` input and invoke the runner on the right loop.

        The QuickJS host call runs on the REPL worker loop, but subagent runnables
        should execute on the parent LangGraph loop when one exists so callbacks,
        context, and async loop affinity match normal tool execution.
        """
        validated = self._validate_task_payload(payload)
        description, subagent_type, label, response_schema = validated

View on GitHub (pinned to a1af029e6e)

Solutions

1. Pass a plain object schema, e.g. `responseSchema: {type: 'object', properties: {answer: {type: 'string'}}}`
2. If you meant to reference a named schema, inline the full object instead
3. Remove `responseSchema` if structured output is not needed

Example fix

// before
await task({ description: 'X', subagentType: 'researcher', responseSchema: 'AnswerSchema' })
// after
await task({ description: 'X', subagentType: 'researcher', responseSchema: { type: 'object', properties: { answer: { type: 'string' } } } })

if (payload.responseSchema !== undefined && payload.responseSchema !== null
    && (typeof payload.responseSchema !== 'object' || Array.isArray(payload.responseSchema))) {
  throw new Error('responseSchema must be an object');
}

function isValidResponseSchema(v) {
  return v === undefined || v === null
    || (typeof v === 'object' && v !== null && !Array.isArray(v));
}

try {
  await task(payload);
} catch (e) {
  if (String(e).includes('responseSchema')) {
    delete payload.responseSchema; // structured output is optional
    return await task(payload);
  }
  throw e;
}

When it happens

Trigger: JS calls `task({..., responseSchema: 'name'})` passing a string, array, boolean, or other non-object value where a JSON object schema is expected.

Common situations: Model-generated JS passing a schema name instead of an inline schema object, marshaling a JS array where an object was intended, or double-encoding the schema as a JSON string.

Understand the failure class

Background: Schema validation failed / invalid input schema: payload rejected because its shape doesn't match the expected schema — this error's family across 28 libraries.

Related errors

• task() requires non-empty string field `description`
• task() requires non-empty string field `subagentType`
• task() field `label` must be a string when provided
• task() requires an active ToolRuntime
• task tool not configured for this eval

AI-assisted analysis of langchain-ai/deepagents@a1af029e6e (2026-08-29). Data as JSON: /api/errors/330c4aa6e09c55b1. Report an issue: GitHub.