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

# Specialization

> Customize the AI persona and reasoning style for your domain

Vrin's specialization feature lets you configure the AI's behavior, tone, and reasoning focus to match your domain.

## specialize()

Set a custom system prompt and optional reasoning parameters.

```python theme={null}
client.specialize(
    custom_prompt="You are a senior financial analyst specializing in tech sector M&A.",
    reasoning_focus=["cross_document_synthesis", "causal_chains"],
    analysis_depth="expert",
)
```

### Parameters

<ParamField body="custom_prompt" type="string" required>
  System prompt that defines the AI's persona, domain expertise, and response style.
</ParamField>

Additional keyword arguments are forwarded to the backend. Common options:

| Parameter         | Type        | Description                            |
| ----------------- | ----------- | -------------------------------------- |
| `reasoning_focus` | `List[str]` | Reasoning strategies to emphasize      |
| `analysis_depth`  | `str`       | `"basic"`, `"detailed"`, or `"expert"` |

### Example specializations

**Legal research:**

```python theme={null}
client.specialize(
    custom_prompt="You are a senior M&A legal partner with 25+ years experience. "
    "Cite specific clauses and precedents. Flag regulatory risks.",
)
```

**Sales enablement:**

```python theme={null}
client.specialize(
    custom_prompt="You are a sales engineer. Answer questions using battle cards, "
    "case studies, and pricing documentation. Be concise and action-oriented.",
)
```

**Technical documentation:**

```python theme={null}
client.specialize(
    custom_prompt="You are a senior software architect. Reference code examples, "
    "API docs, and architecture diagrams. Prefer precise technical language.",
)
```

## get\_specialization()

Retrieve the current specialization settings.

```python theme={null}
settings = client.get_specialization()
print(settings.get("custom_prompt"))
```

### Returns

```json theme={null}
{
  "custom_prompt": "You are a senior financial analyst...",
  "reasoning_focus": ["cross_document_synthesis"],
  "analysis_depth": "expert"
}
```

## Notes

* Specialization is persisted per user -- it applies to all subsequent queries until changed.
* To reset to default behavior, call `specialize()` with a generic prompt.
* Specialization affects the LLM generation step only. Retrieval (graph traversal + vector search) is not modified.