> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getsmelt.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Job Execution

> Running jobs, monitoring progress, and controlling execution

Once you start a job, Smelt processes your rows in parallel for speed.

## Job Statuses

| Status        | Description                     |
| ------------- | ------------------------------- |
| **pending**   | Queued, not yet started         |
| **running**   | Currently processing rows       |
| **paused**    | Temporarily stopped by user     |
| **completed** | All rows processed successfully |
| **failed**    | Error occurred, job stopped     |
| **cancelled** | Manually cancelled by user      |

## What Happens When a Job Starts

1. Status changes to **running**
2. Rows are processed in parallel (5 at a time)
3. Progress updates in real-time
4. Results are saved as they complete
5. Status changes to **completed** when done

## Progress Tracking

While a job runs, you'll see:

| Metric             | Description                      |
| ------------------ | -------------------------------- |
| **Rows processed** | X / total rows completed         |
| **Success count**  | Rows that generated successfully |
| **Failed count**   | Rows that encountered errors     |
| **Progress %**     | Percentage complete              |
| **ETA**            | Estimated time remaining         |

<Note>
  ETA is dynamically calculated based on current processing speed.
</Note>

## Controlling Jobs

### Pause

Click **Pause** to temporarily stop processing:

* Current batch completes before pausing
* No new rows start
* Results so far are saved
* Job can be resumed later

**Use cases:**

* Need to check partial results
* API issues you want to investigate
* Taking a break

### Resume

Click **Resume** to continue a paused job:

* Picks up where it left off
* Previously completed rows are not re-processed
* Continues until all rows done

### Cancel

Click **Cancel** to stop a job permanently:

* Stops all processing
* Completed results are kept
* Cannot be undone
* Remaining rows are not processed

<Warning>
  Cancelled jobs cannot be resumed. Completed results are preserved, but you'd need to create a new job to process remaining rows.
</Warning>

## Parallel Processing

Smelt processes multiple rows simultaneously for speed:

* **Concurrency:** 5 parallel API calls
* **Rate limiting:** Built-in delays to prevent API throttling
* **Retry logic:** 3 attempts per row with exponential backoff

This means:

* Jobs complete faster than sequential processing
* Large jobs (10K+ rows) are still manageable
* Temporary API errors are handled automatically

## What Happens If a Job Fails

If an error stops the job:

1. **Completed rows are saved** — You don't lose progress
2. **Error details available** — Check the job details page
3. **Re-run option** — You can re-run failed rows only

### Common Failure Causes

| Cause                | Solution                             |
| -------------------- | ------------------------------------ |
| API key invalid      | Check Settings > API Keys            |
| Rate limit hit       | Wait and resume, or job auto-retries |
| Usage limit exceeded | Upgrade plan                         |
| Network error        | Check connection, retry              |

## Crash Recovery

If something goes wrong mid-job:

* Results are saved after each row (crash-safe)
* You can see partial results
* Resume or re-run failed rows

<Tip>
  Smelt is designed for large jobs. Processing 10,000+ rows is routine and results are protected throughout.
</Tip>

## Viewing Active Jobs

Go to **Jobs** in the sidebar to see:

* Currently running jobs with live progress
* Recent completed jobs
* Failed jobs that need attention