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

# Fine-tuning

> Shape your assistant's tone and personality and control how it recommends products.

Fine-tuning lets you guide how your assistant sounds and how it suggests products. You set the **Tone**, give **Product advisory** instructions, and choose an **Advisory approach** that fits how your customers like to shop.

## Set the tone

The **Tone** settings shape how the assistant speaks — for example friendly and casual, or precise and professional. The assistant applies your choices across conversations while still answering accurately.

<Steps>
  <Step title="Open Fine-tuning">
    Go to **Assistants** in the [homie platform](https://app.yourhomie.ai), select your assistant, and open **Settings → Fine-tuning**.
  </Step>

  <Step title="Set the tone">
    Choose the **Form of address** (informal or formal) and **Language style** (everyday or technical/professional), toggle **Use emojis**, and add free-text **Additional instructions** for any brand-voice rules you want it to keep.
  </Step>

  <Step title="Add product advisory instructions">
    Under **Product advisory**, tell the assistant what to prioritize when recommending products — for example bestsellers, items on offer, or products that match the visitor's stated needs. The prioritisation of own-brand products can also be specified here.
  </Step>

  <Step title="Choose an advisory approach">
    Select **Needs assessment before recommendation** or **Direct recommendation** (see below).
  </Step>

  <Step title="Test in the Playground">
    Open the [Playground](/en/user-guide/assistants/playground) to confirm the new tone and recommendations feel right.
  </Step>
</Steps>

## Advisory approach

The **Advisory approach** decides how quickly the assistant suggests products. Choose the one that best fits your use case.

### Needs assessment before recommendation

The assistant asks clarifying questions before recommending. Use this when your catalog is broad or products vary a lot, so the assistant can narrow down to the right fit before suggesting anything. It feels more like a guided, consultative conversation.

### Direct recommendation

The assistant recommends relevant products right away, based on what the visitor said. Use this when shoppers want fast suggestions or your catalog is focused, so they reach a product without extra back-and-forth.

<Tip>
  Choose **Needs assessment before recommendation** for considered, high-variety purchases and **Direct recommendation** for quick, impulse-friendly shopping. Re-test in the Playground after switching.
</Tip>

## Next steps

<CardGroup cols={2}>
  <Card title="Test your changes" icon="flask" href="/en/user-guide/assistants/playground">
    Try the new tone and recommendations live.
  </Card>

  <Card title="Improve product answers" icon="book" href="/en/user-guide/knowledge/product-data">
    Make sure product data is complete and accurate.
  </Card>
</CardGroup>