# Chalk CLI Reference: Evaluations
source: https://docs.chalk.ai/cli#evaluations

Inspect evaluations, trigger evaluation runs, and review the scores those runs produced.

## `chalk eval list`
source: https://docs.chalk.ai/cli?command=eval_list

```bash
chalk eval list [flags]
```

Use the 'chalk eval list' command to view the evaluations defined in your environment, along with the average score of each evaluation's most recent scored run.

Use --json to get the full output, including the score summaries, as JSON.

**Flags**

- `--limit <limit>` (int, default `50`)
  Maximum number of evaluations to return per page.
- `--all <all>` (bool, default `false`)
  Fetch every evaluation by paginating through all results.

**Example**

```
$ chalk eval list
$ chalk eval list --limit 10
$ chalk eval list --all --json > evaluations.json
```

## `chalk eval get`
source: https://docs.chalk.ai/cli?command=eval_get

```bash
chalk eval get <evaluation-id>
```

Use the 'chalk eval get' command to retrieve the definition of an evaluation: the dataset revision it runs against, the task function it evaluates, and the scorers applied to its output.

**Arguments**

- `evaluation-id` (string)
  The ID of the evaluation to fetch.

**Example**

```
$ chalk eval get evl_01hxyz
$ chalk eval get evl_01hxyz --json
```

## `chalk eval archive`
source: https://docs.chalk.ai/cli?command=eval_archive

```bash
chalk eval archive <evaluation-id>
```

Archive the evaluation in the current environment. Does not cancel running work or delete datasets. Hides the evaluation and its runs from lists.

**Arguments**

- `evaluation-id` (string)
  The ID of the evaluation to archive.

## `chalk eval suite archive`
source: https://docs.chalk.ai/cli?command=eval_suite_archive

```bash
chalk eval suite archive <suite-id>
```

Archive the suite and hide its evaluations and runs from lists. Does not cancel running work or delete datasets. The Default suite cannot be archived.

**Arguments**

- `suite-id` (string)
  The ID of the evaluation suite to archive.

## `chalk eval trigger`
source: https://docs.chalk.ai/cli?command=eval_trigger

```bash
chalk eval trigger <evaluation-id>
```

Use the 'chalk eval trigger' command to start a new run of an existing evaluation. The run executes the evaluation's task function over its pinned dataset revision and scores the output.

The command returns as soon as the run is created unless '--await' is passed, in which case it polls until the run succeeds, fails, or is canceled.

**Arguments**

- `evaluation-id` (string)
  The ID of the evaluation to run.

**Flags**

- `--metadata <json>` (string)
  JSON object to attach to the evaluation run as metadata, e.g. `'{"trigger":"ci"}'`.
- `--await <await>` (bool, default `false`)
  If `true`, poll the run until it reaches a terminal status before exiting.
- `--timeout <duration>` (string, default `1h`)
  How long to wait when `--await` is set, as a Go duration, e.g. `'30m'`.

**Example**

```
$ chalk eval trigger evl_01hxyz
$ chalk eval trigger evl_01hxyz --metadata '{"trigger":"ci","commit":"abc123"}'
$ chalk eval trigger evl_01hxyz --await --timeout 30m
```

## `chalk eval scores`
source: https://docs.chalk.ai/cli?command=eval_scores

```bash
chalk eval scores [evaluation-id]
```

Use the 'chalk eval scores' command to see how evaluation runs scored: the average score of each run and, with '--run', the per-scorer averages, per-metric values, and the percentile distribution behind them.

Pass an evaluation ID to summarize only that evaluation's runs. Runs that have not been scored yet are listed with an empty score rather than omitted.

**Arguments**

- `evaluation-id` (string)
  Only summarize runs of this evaluation.

**Flags**

- `--run <evaluation-run-id>` (string)
  Show the full per-scorer and per-metric breakdown for this run only.
- `--limit <limit>` (int, default `50`)
  Maximum number of runs to summarize per page.
- `--all <all>` (bool, default `false`)
  Summarize every matching run by paginating through all results.

**Example**

```
$ chalk eval scores
$ chalk eval scores evl_01hxyz
$ chalk eval scores evl_01hxyz --run evr_01hxyz
$ chalk eval scores evl_01hxyz --all --json > scores.json
```

## `chalk eval run list`
source: https://docs.chalk.ai/cli?command=eval_run_list

```bash
chalk eval run list [flags]
```

Use the 'chalk eval run list' command to list evaluation runs in your environment, optionally narrowed to a single evaluation.

Each row shows the run's status and, once it has been scored, its average score. Use 'chalk eval scores' for the per-scorer breakdown.

**Flags**

- `--evaluation <evaluation-id>` (string)
  Only list runs of this evaluation.
- `--limit <limit>` (int, default `50`)
  Maximum number of runs to return per page.
- `--all <all>` (bool, default `false`)
  Fetch every matching run by paginating through all results.

**Example**

```
$ chalk eval run list
$ chalk eval run list --evaluation evl_01hxyz
$ chalk eval run list --evaluation evl_01hxyz --all --json > runs.json
```

## `chalk eval run get`
source: https://docs.chalk.ai/cli?command=eval_run_get

```bash
chalk eval run get <evaluation-run-id>
```

Use the 'chalk eval run get' command to retrieve the status of an evaluation run, the dataset revision holding its output, and the error message if it failed.

Fetching a run also reconciles it against the underlying query, so this is the command to poll while a run is in progress.

**Arguments**

- `evaluation-run-id` (string)
  The ID of the evaluation run to fetch.

**Example**

```
$ chalk eval run get evr_01hxyz
$ chalk eval run get evr_01hxyz --json
```

## `chalk eval run archive`
source: https://docs.chalk.ai/cli?command=eval_run_archive

```bash
chalk eval run archive <evaluation-run-id>
```

Archive the evaluation run in the current environment. Does not cancel running work or delete datasets. Clears its baseline flag.

**Arguments**

- `evaluation-run-id` (string)
  The ID of the evaluation run to archive.

## `chalk eval run set_baseline`
source: https://docs.chalk.ai/cli?command=eval_run_set_baseline

```bash
chalk eval run set-baseline <evaluation-run-id>
```

Use the 'chalk eval run set-baseline' command to choose the run that an evaluation's other runs are compared against.

Exactly one run per evaluation is the baseline, so this clears the flag on whichever run held it before.

**Arguments**

- `evaluation-run-id` (string)
  The ID of the evaluation run to make the baseline.

**Example**

```
$ chalk eval run set-baseline evr_01hxyz
```
