How to Add a Knowledge Base to the Tess AI Chat: Use Files and Websites Within Your Chat 🗒️

Did you know it's possible to make the Tess AI chat even more powerful and personalized? By adding a knowledge base, you can provide specific context for your conversations, significantly improving the relevance and accuracy of the responses. In this guide, we'll show you how to do this simply and effectively.

What is a knowledge base in the Tess AI chat?

A knowledge base is a set of information that you add to the Tess AI chat so it can use it as a reference during conversations. This allows the AI to access specific and relevant data for your needs.

Supported file types

Tess AI supports various file formats for your knowledge base:

• PDF
• Audio
• Google Sheets
• Web pages
• Excel
• CSV

How to add a knowledge base

Adding a knowledge base is a simple process:

1. Open the Tess AI chat.
2. Locate the paperclip icon next to the command field.
3. Click the icon to open the options.
4. Select "Add knowledge base."
5. Choose the type of file you want to add.
6. Upload the desired file.

Limits and recommendations

When adding files to your knowledge base, remember:

• The maximum size per file is 200 MB.
• You can add files until you reach the limit of 200k Tokens.

How to use the knowledge base in the chat

To use the information from your knowledge base, ask Tess to use the information from your file by mentioning its name, for example: "Use the PDF 'Annual_Report_2023' to generate a summary of the main results."

Removing files from the knowledge base

Keeping your knowledge base organized is crucial for the efficient performance of the Tess AI chat. To remove files that are no longer needed:

1. Open the Tess AI chat.
2. Locate the paperclip icon next to the writing field.
3. Click on the option "Open the knowledge base."
4. Hover over the file you want to delete.
5. Click the trash can icon that appears.

Tips to optimize your knowledge base

• Organize your files: Use descriptive names for easy reference.
• Stay within limits: Avoid exceeding 200 MB per file.
• Combine different types of files: Use PDFs for long documents, audios for transcriptions, and web pages for dynamic information.

What if I need a very large knowledge base? Understanding RAG

In that case, we recommend using the RAG methodology, a feature that allows you to expand the knowledge of your chat templates beyond what the models support in content volume, opening possibilities to create increasingly trained chats.

LLMs are trained on massive datasets, allowing them to generate text, translate languages, and answer questions impressively. However, this knowledge is static to a certain extent, meaning if the LLM has a context window of 200k tokens, the content you will use to train it in the Knowledge Base cannot be larger than that.

RAG breaks this barrier, allowing your chats to access and read extra information, in addition to the Knowledge Base attachments, complementing the provided context and bringing more complete answers.

To do this, simply select the context mode (Context Mode *) as RAG, as shown in the example below.

In practice, RAG reads through associations and mentions of certain topics. For example, you can insert a 50-page file, but in RAG, the AI will only search for what is related to your question. Due to this optimized work, RAG allows increasing the context window of that LLM.

What is the difference in credit consumption?

When adding a knowledge base, there is a small processing cost that will consume your credits only once, exactly to access your loaded base.

And generally, it is very little, don't worry. Without additional base and unlimited model, you don't spend any of your credits!

Conclusion

By adding a knowledge base to the Tess AI chat, you are significantly expanding its capabilities. Try adding different types of files and see how this can transform your interactions with the AI.

Remember: the quality of your knowledge base directly influences the quality of Tess AI's responses. Invest time in creating a robust and relevant base, and you will see impressive results in your conversations.