> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://www.comet.com/docs/opik/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://www.comet.com/docs/opik/_mcp/server.

# Dialogue helpfulness

> Measure how helpful an assistant reply is within a dialogue

# Dialogue Helpfulness Judge

`DialogueHelpfulnessJudge` inspects the latest assistant reply in the context of preceding turns. It rewards responses that acknowledge the user’s request, use the available context, and offer actionable guidance.

**`Scoring a support reply`**

```python title="Scoring a support reply"
from opik.evaluation.metrics import DialogueHelpfulnessJudge

turns = """USER: My VPN disconnects every 5 minutes.\nASSISTANT: Try reinstalling the client.\nUSER: I already did.\n"""

metric = DialogueHelpfulnessJudge()
score = metric.score(
    input=turns,
    output="Can you send logs? I'll escalate to network engineering.",
)

print(score.value)
print(score.reason)
```

## Inputs

| Argument       | Type         | Required | Description                                                 |                                |
| -------------- | ------------ | -------- | ----------------------------------------------------------- | ------------------------------ |
| `input`        | `str`        | Optional | Conversation history (alternating USER / ASSISTANT blocks). |                                |
| `conversation` | `list[dict]` | Optional | Structured turns (`{"role": "user", "content": "..."}`      | `{"role": "assistant", ...}`). |
| `output`       | `str`        | **Yes**  | Latest assistant reply to score.                            |                                |

## Configuration

| Parameter      | Default      | Notes                                                          |
| -------------- | ------------ | -------------------------------------------------------------- |
| `model`        | `gpt-5-nano` | Switch to a larger evaluator for complex enterprise workflows. |
| `temperature`  | `0.0`        | Use low temperature for reproducible benchmarks.               |
| `track`        | `True`       | Record the evaluation in Opik.                                 |
| `project_name` | `None`       | Set when routing results to a different project.               |

Integrate this judge into regression suites to catch regressions after prompt changes or upgrades to your assistant model.