Skip to main content

Step 2: Get answers on your documents

Upload a file to an AI agent and get immediate answers​

Agent Studio is the quickest way to start building a custom AI agent. In this tutorial, you'll build on the skills you developed in step 1 of this series by creating a knowledge agent with file upload and chat capabilities.

What you'll build​

  • An AI agent connected to a knowledge base that can provide answers to questions on an uploaded file.

What you'll need​

  • This tutorial builds on the skills developed in step 1 of this series, where you create your first AI agent. If you haven't completed this tutorial, please do so before continuing.
  • A Squid account in the Squid Console

Create a New Agent​

  1. In the Squid Console, navigate to the Agent Studio tab.

  2. Click Create New Agent to add an agent. A window will appear asking for an Agent ID and description.

    • Agent ID: knowledge-agent.
    • Description: This is an agent for Squid AI's document and knowledge base tutorial

  3. Click Create to create the agent.

After creating the agent, you will be redirected to the Overview tab of your new agent.

Give the Agent Instructions​

Providing the agent with custom instructions will give it a unique persona. Because this is a knowledge agent that will respond to user prompts, provide instructions that guide the agent to focus on information in a provided file:

Your task is to answer user prompts based on the file provided to you. 
Only answer questions that can be answered using the information in the file.

Create the Knowledge Base​

Squid AI's knowledge base feature provides additional information to the agent when answering questions. Connecting a knowledge base to an agent enables the agent to provide relevant answers on specific topics that may not be part of the underlying AI model. This allows for custom AI agents that have access to domain-specific knowledge and company information.

To create a knowledge base for your AI agent, follow these steps:

  1. Navigate back to the home page of Agent Studio. The home page contains three different tabs – Agents, Knowledge Bases, and Settings. Click on the Knowledge Bases tab.

  2. Click Create New Knowledge Base to add a new knowledge base. A window will appear asking for a Knowledge Base ID and a description of when the agent should consult the knowledge base.

    • Knowledge Base ID: earnings-calls
    • Description: Use this knowledge base when asked to provide information on earnings calls
    • Embedding Model: You can leave the default embedding model selected for this tutorial.

    Note: If you would prefer to use your own file rather than the sample provided in this tutorial, be sure to provide an appropriate ID and description for your use case.

    After providing the details, click Create to create the knowledge base.

  3. You will then be redirected to the earnings-calls knowledge base. Click on Add Knowledge in the top right corner of the page. A window will appear asking you to upload a file.

    This tutorial provides Microsoft's Q3 2025 earnings call transcript as a sample file. You can download the file here.

    Next, drag and drop the msft_earnings_transcript_25_q3.pdf file into the upload window, or click on the window to browse your computer for the file. After selecting the file, click Add. Squid AI will then process, optimize, and add the file to the knowledge base.

  4. Next, navigate back to the Agents tab within Agent Studio, and click on your knowledge-agent to open the agent's Overview tab.

  5. In the Overview tab of your agent, click on the Add Abilities button.

  6. Find the Knowledge Bases section of the sidebar. Click on Your Knowledge Bases to see the list of knowledge bases you have created. Click on the + next to the earnings-calls knowledge base to add it to your agent. A window will appear asking you to provide a description of what information this knowledge base contains. Provide the following description:

    • Description: This knowledge base contains information on earnings calls

    After providing the description, click Connect Knowledge Base.


Nice one! Your agent can now access the information in the uploaded file!

Test Your Agent​

After adding the knowledge base, navigate to the Test Agent tab of the agent. This tab allows you to interact with your agent and see how it responds to user input.

Test the agent by asking a few questions about the uploaded file:

What was Microsoft's reported revenue?

Did Microsoft's on-premises server revenue increase or decrease year-over-year?

How many MAU (Monthly Active Users) does Xbox have?

What new AI features were announced for Microsoft 365?

What color is the sky?

These prompts will trigger the agent to consult the knowledge base, providing the most relevant information found in the document. The agent will also use its instructions to only respond with information found in the file. Therefore, when asked about the color of the sky, the agent should respond that it does not have that information.

Next Steps​

Congratulations! You've successfully created an AI knowledge agent. You now have a basic understanding of how to create and customize a helpful agent.

  • Continue with the next tutorial in the Agent Studio onboarding series: Create a PDF Agent.

  • Want to build a highly customized AI agent workflow? Build with Squid's SDKs to create a custom agent, or check out an AI tutorial for a hands-on learning experience. For technical developers only.

  • Have a question? Please reach out to us! We are happy to help you get started.