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

# Agent Training

> Equip your agent with Knowledge Base documents, external API tools, and cross-call memory

The **Training** tab gives your agent the specific knowledge and real-time capabilities it needs to resolve customer inquiries. Rather than relying solely on general model knowledge, training allows your agent to consult your company documents and take real actions mid-call.

![Training Tab in Vaani Dashboard](https://mintcdn.com/vaaniai/1z97x591P_yo5Lyb/images/screenshots/training-agent.png?fit=max\&auto=format\&n=1z97x591P_yo5Lyb\&q=85\&s=9dd5e5e257214bbbd3870d8c8004b5ab)

## Video Walkthrough

[https://www.loom.com/embed/74edda24a2eb4e5e995fc7358c846983](https://www.loom.com/embed/74edda24a2eb4e5e995fc7358c846983)

***

## 1. Knowledge Base (Retrieval-Augmented Generation / RAG)

Upload your company's product catalogs, return policies, price charts, and FAQs directly to your agent.

> **Document Upload**
>
> Upload PDF, DOCX, TXT, CSV, or Markdown manuals. Vaani extracts text and indexes vectors automatically.

> **Website Crawling**
>
> Enter your documentation or help center URL. Vaani recursively indexes your help pages.

> **Live Knowledge Retrieval**
>
> During calls, relevant chunks are retrieved in \~100ms and supplied to the model as context.

### Best Practices for Knowledge Documents

* **Keep files organized**: Use clear section headings (`## Pricing`, `## Refund Policy`).
* **Include exact numbers**: Provide exact timelines, fees, and office hours.
* **Instruct fallback behavior**: In the system prompt, specify: *"If the information is not found in your knowledge base, politely state that you'll have a specialist follow up."*

***

## 2. Tools & Actions (Function Calling)

Empower your agent to perform actions while talking to the caller - such as looking up an order, sending an OTP, checking a calendar, or submitting a lead to your CRM.

### Adding a Tool via cURL

You can define a new tool in seconds by pasting a standard `cURL` request:

```bash theme={null}
curl -X POST https://api.yourstudio.com/v1/appointments/check \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"date": "2026-10-25", "time": "14:00"}'
```

Vaani parses the endpoint, headers, and request body parameters automatically:

| Parameter | Type | Description |
| - | - | - |
| `date` | `string` | The requested consultation date (YYYY-MM-DD) |
| `time` | `string` | The requested time slot in IST |

### Natural Language Tool Triggering

Provide a concise description of when the tool should be executed:

> *"Trigger this tool whenever the caller asks to book an appointment or check designer availability for a specific day."*

***

## 3. Context & Memory

Ensure callers never have to repeat themselves across different calls or departments:

**Previous Calls Memory**

Remembers caller preferences, previous issues, and unfinished inquiries tied to their phone number. When the customer calls back, the agent greets them personally:

> *"Welcome back, Priya! Are you calling regarding the 3BHK interior estimate we discussed on Tuesday?"*

**Cross-Agent Memory**

Shares conversational context across different agents in your organization (e.g. Sales Agent handoff to Support Agent).

***

## Next Steps

> **[Configure Experience](https://docs.vaanivoice.ai/getting-started/agent-config/experience.md)**
>
> Tune ambient background noise, speech delays, and interruption thresholds

> **[Post-Call Analysis](https://docs.vaanivoice.ai/getting-started/agent-config/analysis.md)**
>
> Define structured data extraction and automated QA evaluation


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.