> For the complete documentation index, see [llms.txt](https://docs.layerlens.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.layerlens.ai/more-in-this-section-9/judges-2.md).

# Judges

client.judges — CRUD for evaluation judges, with version-on-update semantics.

## Create

`name` and `evaluation_goal` are required. `model_id` is optional — if omitted, a default judging model is used.

```python
from layerlens import Stratix

client = Stratix()

# Optionally fetch a model to use as the judge's LLM
models = client.models.get(type="public", name="gpt-4o")
model = models[0]

judge = client.judges.create(
 name="Code Quality Judge",
 evaluation_goal="Evaluate the quality of code output including correctness, readability, and style",
 model_id=model.id,
)
print(f"Created judge {judge.id}: {judge.name}")
```

## Get

```python
judge = client.judges.get(judge_id)
print(f"Judge: {judge.name}, version: {judge.version}")
```

## List

`get_many()` returns a `JudgesResponse` with `.judges`, `.total_count`:

```python
response = client.judges.get_many()
print(f"Found {response.total_count} judges")
for j in response.judges:
 print(f" - {j.name} (v{j.version}, {j.run_count} runs)")
```

## Update — creates a new version

Updating a judge **creates a new version**. The prior version remains accessible for already-attached evaluations:

```python
updated = client.judges.update(
 judge.id,
 name="Updated Code Quality Judge",
 evaluation_goal="Evaluate code output for correctness, readability, style, and security",
)
print(f"Updated judge {updated.id}, version {updated.version}")
```

## Delete

```python
deleted = client.judges.delete(judge.id)
print(f"Deleted judge {deleted.id}")
```

## Async

```python
from layerlens import AsyncStratix

async def example():
 client = AsyncStratix()
 judge = await client.judges.create(name="...", evaluation_goal="...")
 response = await client.judges.get_many()
```

## Versioning model

* Each `update()` call creates a new `JudgeVersion` linked to the parent judge
* The judge's `.version` field reflects the latest version
* Already-running evaluations stay pinned to the version they started with
* Use `JudgeSnapshot` for point-in-time references in audit trails

## Source samples

| Sample                                    | What it shows                                      |
| ----------------------------------------- | -------------------------------------------------- |
| `samples/core/create_judge.py`            | Full CRUD lifecycle                                |
| `samples/core/judge_creation_and_test.py` | Build a custom PII detection judge with validation |
| `samples/core/judge_optimization.py`      | Estimate, run, apply judge optimizations           |
| `samples/core/trace_evaluation.py`        | Apply judge to traces                              |
| `samples/industry/healthcare_clinical.py` | Domain-tuned clinical judge                        |

## See also

* [Concept: Judges](/more-in-this-section-9/judges-2.md)
* [Judge optimizations (GEPA)](/more-in-this-section-9/judge-optimizations-1.md)
* [Tutorial 2: Build your first judge](/8.-evaluate-score-the-outputs/02-first-judge.md)
* [Bootstrap a judge before GEPA](/more-in-this-section-6/bootstrap-judges.md)