Handle Large Task Results

When a Task Studio task returns a very large payload (for example, a big list of contacts or rows pulled from an external API), the result can exceed the maximum size a task run is allowed to return inline. When that happens the run still completes successfully, but the oversized payload can’t be returned or previewed directly, and you’ll see a message like:

This task returned too much data to display or pass along directly.

This guide shows the two ways to work around that limit by persisting the full output as a reusable asset.

Option 1: Save the result as an asset (no code changes)

The quickest fix requires no changes to the task itself — just ask the agent to re-run the task and save its result as an asset. For example:

Re-run that task and save the result as an asset.

The agent re-runs the task, stores the complete output as a tool_result asset, and returns an asset_id instead of the raw payload. That asset_id can then be handed to other tools — for instance, to bulk-load the rows into a database — without the data ever being re-typed or truncated.

Option 2: Update the task to write an asset itself

For a task that reliably produces large outputs, the durable fix is to have the task create an asset and return a reference to it rather than returning the raw data. You can ask the agent to update the script for you, or edit it directly.

Instead of returning a large array:

1
def main(company: str) -> dict:
2
    contacts = fetch_all_contacts(company)  # potentially thousands of rows
3
    # ❌ Returning the full array inline can exceed the result-size limit
4
    return {"contacts": contacts}

…create an asset from the SDK and return its id plus a small summary:

1
from athena import Athena
2
3
4
def main(company: str, api_key: str) -> dict:
5
    client = Athena(api_key=api_key)
6
    contacts = fetch_all_contacts(company)
7
8
    # ✅ Persist the full dataset as a spreadsheet asset the agent can reuse
9
    asset = client.assets.create(
10
        asset_type="spreadsheet",
11
        title=f"{company} contacts",
12
    )
13
    populate_spreadsheet(client, asset.asset_id, contacts)
14
15
    # Return a small, inline-safe reference instead of the full payload
16
    return {
17
        "asset_id": asset.asset_id,
18
        "row_count": len(contacts),
19
        "text": f"Saved {len(contacts)} contacts for {company} to an asset.",
20
    }

The task now returns a compact response that always fits inline, while the full dataset lives in an asset that downstream tools and the agent can open and reuse later.

See Create Assets for the full asset-creation API and Build with Agents for how tasks and agents work together.