---
title: "Agents, workers and repairs"
description: "Jason builds and repairs the pipeline. Workers execute its steps to produce the dataset."
url: "https://www.jsonify.com/docs/data-agents"
---

[Documentation](https://www.jsonify.com/docs/)[What is Jsonify?](https://www.jsonify.com/docs/what-is-jsonify)[How pipelines work](https://www.jsonify.com/docs/how-jsonify-works)[Radar or Benchmark? Enterprise](https://www.jsonify.com/docs/radar-vs-benchmark) [Agents, workers and repairs](https://www.jsonify.com/docs/data-agents)[Row pricing and managed service](https://www.jsonify.com/docs/pricing-model)[Connect your agent →](https://www.jsonify.com/connect)

# Agents, workers and repairs

Jason builds and repairs the pipeline. Workers execute its steps to produce the dataset.

## Jason: build and maintain

Jason is the agent you describe the data task to. It explores sources, builds extraction code and assembles the pipeline. When an extraction step stops working, Jason can investigate the failure and produce a revised implementation.

This is different from asking an agent to return a one-off answer. The result is a pipeline that can run again.

## Workers: execute the steps

Workers carry out the pipeline’s collection and processing steps. Independent work can run in parallel; dependent work follows the sequence defined by the pipeline.

A restaurant-menu pipeline might find venue websites and delivery listings in parallel, extract menu data from each, then merge and verify the records before publishing.

## Datasets: the structured result

A dataset contains the fields your pipeline produces. Useful fields often include the source URL, collection time, units and currency alongside the main values. They help you interpret changes and compare records correctly.

Verification can check missing values, types, duplicates and other expectations specific to the task. A successful fetch alone does not establish that the extracted result is correct.

## What automatic repair means

Suppose a retailer moves its price field. A pipeline step can fail because the source no longer matches the implementation. Jason diagnoses the failure, changes the extraction code and verifies the revised step against a sample before a new revision becomes active.

The pipeline then runs with the repaired code. Changes to code and subsequent runs can be tracked through revisions and run history.

## What it does not mean

Sources may become unavailable, and some fixes need review. A repair must pass verification before the revised pipeline becomes active. If it cannot be verified, the issue needs further investigation.

For ongoing operational support and a managed scope, see [Enterprise](https://www.jsonify.com/enterprise).

[All documentation](https://www.jsonify.com/docs/)[Build it →](https://factory.jsonify.com/setup)