> For the complete documentation index, see [llms.txt](https://docs.suvvy.ai/en/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.suvvy.ai/en/main-settings/prompts/audit-of-instructions.md).

# Audit of instructions

Manual audit is an AI check that helps improve the bot's existing configuration. In basic mode, it only adjusts the manual. In advanced mode, it adjusts the manual and files.

Auditing the instructions helps automatically improve the existing bot settings.

The system analyzes:

* the system user manual
* connected functions
* knowledge base files
* general structure of bot settings

Suvvy then offers a more accurate and logical version of the instructions, and in advanced mode, it also helps you organize your knowledge base.

Go to the Instructions section and click Audit Instructions

<figure><img src="https://2542240793-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnmK9dnkx224HgS0VSB5J%2Fuploads%2FhnUfN8V1iLnnmvXMjkfo%2Fimage.png?alt=media&amp;token=c315c5c2-fb06-47bc-977e-83e8f118be60" alt=""><figcaption></figcaption></figure>

{% hint style="danger" %}
Please note that auditing the instructions and Knowledge Base files **requires tokens**.

Using this tool is charged at the platform's **standard pricing**.
{% endhint %}

#### We have 2 modes of operation:

With and without formation of a knowledge base file:

<figure><img src="https://2542240793-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnmK9dnkx224HgS0VSB5J%2Fuploads%2F4Ds3iR5hupt96S2cuIeB%2Fimage.png?alt=media&amp;token=3f39dd36-588f-43a7-9786-a23668ebd628" alt=""><figcaption></figcaption></figure>

Auditing the instructions can take up to 10 minutes

<figure><img src="https://2542240793-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnmK9dnkx224HgS0VSB5J%2Fuploads%2FAF30xhSizhcMKXu8GNzz%2Fimage.png?alt=media&amp;token=f6037728-4231-472a-80e6-574d8f9208ae" alt=""><figcaption></figcaption></figure>

### Mode 1 Audit instructions (select not to create knowledge base files)

This is the basic testing mode.

**The system receives current instructions, knowledge base files, and other functions and checks:**

* structural logic;
* completeness of key sections;
* duplication of meanings;
* contradictions;
* redundant or unclear wording;
* readability and suitability of the instructions for LLM;
* using correct function names

**After Suvvy's analysis:**

* оrganizes sections;
* makes wording clearer;
* removes unnecessary repetitions;
* makes the instructions more coherent and suitable for the bot's work

<figure><img src="https://2542240793-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnmK9dnkx224HgS0VSB5J%2Fuploads%2FManSOGFdjQjcWQJ2cRLD%2Fimage.png?alt=media&amp;token=44bc8448-4070-4932-bb1b-8abdf7cfa944" alt=""><figcaption></figcaption></figure>

### Mode 2 Audit instructions + Generation of knowledge base files (select  create knowledge base files)

This is an advanced mode.

In addition to instructions, the AI ​​immediately identifies what can be transferred to the Knowledge Base and creates appropriate files for direct questions.

If necessary, it structures and defines information in the Knowledge Base files.

<figure><img src="https://2542240793-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnmK9dnkx224HgS0VSB5J%2Fuploads%2F6S6JCB1syHn2vxIxerkt%2Fimage.png?alt=media&amp;token=4e41ea5c-0052-4d42-ad3d-b0d0455bdae2" alt=""><figcaption></figcaption></figure>

<figure><img src="https://2542240793-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnmK9dnkx224HgS0VSB5J%2Fuploads%2FQCzWXv1uUswFm3SZKAtn%2Fimage.png?alt=media&amp;token=e1748da3-aaf9-4c09-b3c9-78c3f69c8451" alt=""><figcaption></figcaption></figure>

How does an audit work?

{% stepper %}
{% step %}

### Step 1. Launch

Click Audit Instructions
{% endstep %}

{% step %}

### Step 2. Analysis

AI analyzes the bot's current configuration and prepares an improved version.
{% endstep %}

{% step %}

### Step 3. View results

Before use, check:

* Have important constraints been lost?
* Has the desired tone of communication changed?
* Have the functions and structure of the knowledge base been correctly taken into account?
* Have critical formulations for your case been removed?
  {% endstep %}

{% step %}

### Step 4. Application

If you are satisfied with the result, click Apply.

{% hint style="warning" %}
We recommend that you additionally check the files in Direct Questions
{% endhint %}
{% endstep %}
{% endstepper %}