Connect Any Data Source To Build the AI Agent You Need

Thinkstack turns your data-FAQs, PDFs, CSVs, URLs, meeting notes via Zapier, and Notion—into answers and solutions. Your team works faster, and your customers get what they need in no time

Multiple Knowledge Sources for Training

Files

Upload data from .pdf, .docx, .txt, or .md files, like manuals and training guides.

URLs

Keep your AI Agent updated with live information from your website or portals.

Q&A inputs

Add FAQs to help your AI Agent handle customer questions easily.

CSVs

Find patterns and trends in spreadsheets to make better decisions.

Notion

Sync your team’s tasks and knowledge for easy access.

How does this feature of multiple data sources integration come handy?

Quick answers, less hassle

Get instant replies to questions like:

  • “What’s the return policy?”
  • “What were last week’s meeting takeaways?”

Make it personal

Your AI Agent remembers customer and team needs, giving better, customized answers.

Easy access to resources

From product guides to meeting notes, find everything you need right away.

Help customers and teams help themselves

Let your AI Agent power FAQs and guides for smooth self-service.

Get accurate answers every time

With up-to-date knowledge, your AI Agent gives answers you can trust..

Ready to work smarter with your knowledge?  

Let Thinkstack AI Agents turn your data into real answers and solutions.

Build your AI chatbot for free No Credit Card Required

Frequently asked questions

What is a Data Source?

minus-circle plus-circle

A Data Source is any repository of information used to train your chatbot. This can include files (like PDFs, Word documents, text and markdown files), URLs, Q&A pairs, CSVs or integrations with platforms like Notion or Zapier. Data sources provide the content that the bot uses to generate responses.

What Are Knowledge Base Characters?

minus-circle plus-circle

Knowledge base characters are the cumulative count of all text characters within the Data Source (files, Q&A entries, web pages, etc.) that the bot is trained on.

What is the "Crawler Type" option, and what are its available types?

minus-circle plus-circle

The "Crawler Type" option determines how data is extracted from a website during crawling. The available types are:

1. Auto-detect: Automatically selects the best crawling method based on the website's structure.

2. HTML: Extracts data from static HTML content.

3. JavaScript: Extracts data from dynamically rendered web pages.

When should I choose a specific Crawler Type (HTML, JavaScript, or Auto-detect)?

minus-circle plus-circle

Use HTML for websites with static content.

Choose JavaScript for dynamically rendered web pages or single-page applications.

Select Auto-detect to let the system determine the best method automatically.

Can I train my chatbot with multiple URLs?

minus-circle plus-circle

Yes, you can. The number of links that you can use to train your chatbot depends on your subscription plan. Check our pricing page for more information.

Can I exclude specific pages from training?

minus-circle plus-circle

Yes, you can exclude specific pages by using glob patterns or by unchecking them from the list of URLs after the crawling process is complete.

What should I do after the crawling process is complete?

minus-circle plus-circle

Review the list of URLs, uncheck any pages you don’t want to include, and click "Train now" to begin training your chatbot. The training time will depend on the number of pages.

What are the files that I can upload to train a chatbot?

minus-circle plus-circle

You can upload a variety of file types to train your chatbot, including PDFs, Word documents (.doc, .docx), text files (.txt), and Markdown files (.md). Ensure that the files are within 50MB, well-structured, and contain relevant, clear content for effective training.

What Q&A I should add to train the chatbot?

minus-circle plus-circle

Sometimes, users might ask a question that the chatbot cannot answer. To address this, observe such questions and add them as a Question with appropriate responses Here are some types of Q&A you can include:

Frequently Asked Questions (FAQs): Add common queries that users are likely to ask.

Specific Business Information: Product details, policies, or services, ongoing offers, and benefits.

Customer Support Scenarios: Provide responses to common issues or complaints to assist users effectively.

What should I do if the chatbot gives incorrect answers?

minus-circle plus-circle

Use the "Chats" section to access the chat history, revise the chatbot’s incorrect responses, and save the corrected response as a Q&A for future use.

Can I configure Q&A responses as clickable suggestions?

minus-circle plus-circle

Yes, you can set specific Q&A responses to appear as suggestion prompts, allowing users to get answers with a single click.

Can I train the chatbot with the Notion Document URL?

minus-circle plus-circle

No, you cannot directly train the chatbot with a Notion document URL. However, we offer Notion integration; you can connect your Notion account and select specific pages or databases for training.