Skip to main content

Work with AI

AI collaboration overview with whiteboard, cards, and PDF

TL;DR​

In Heptabase, you can open Chat on a whiteboard to chat with the latest AI models. By adding cards, sections, whiteboards, or content like PDFs, videos, and journals to the chat context, you enable AI to respond based on the content you select. You can also edit and annotate important messages before placing them back on the whiteboard, where they can be visually organized alongside your existing notes for centralized management of topic-related discussions. By working with AI, you can transform linear chat records into a more complete knowledge structure, continuously building deeper understanding of knowledge topics.

Introduction​

In the AI era, people use prominent AI apps like ChatGPT, Google Gemini, and Claude to learn and research topics they care about. However:

  1. When topics are complex and involve connections between many different concepts, you may wish to extract key knowledge points provided by AI from linear conversation threads and place them on a large whiteboard for easier research and review.

  2. When you ask AI questions, you may want AI to read notes you've previously written or saved knowledge content, and provide valuable responses based on this content.

These are exactly what Heptabase's AI features excel at. In this article, I'll walk you through methods for using AI in Heptabase to accelerate learning and research, and share common use cases and workflows for different types of knowledge sources.

Overview​

Generally, collaborating with AI in Heptabase includes the following core actions:

  1. Open Chat in a topic whiteboard to centrally manage conversations related to that topic.

  2. Use the add button "+" or "@" mention to provide the whiteboard and specific content to AI as context.

  3. Right-click on PDFs, videos, journals, and cards on the whiteboard to add them to Chat, and have conversations based on these sources.

  4. Drag valuable messages to the whiteboard or copy them to cards for editing and adding your own annotations.

  5. Hover over any card on the whiteboard to reveal AI action buttons, and use it to perform default or custom AI actions.

  6. Use sections, sub-whiteboards, or arrows to gradually integrate AI-assisted understanding into your knowledge graph.

Researching a Topic with AI​

Research a topic

You can find the Research a topic button in the left sidebar. Clicking it opens an interface where you can upload various types of content, including PDFs, YouTube links or playlists, .docx files, plain text documents, and images. After uploading, simply click Start research on a new whiteboard, and the system will add all uploaded content to a new whiteboard. It will automatically parse PDF content and download YouTube video transcripts. Then you can:

  • Start asking AI any questions about the whiteboard
  • AI responses will include citation links to the original sources, pointing to specific paragraph blocks or timestamps—just click to jump to the corresponding location
  • If you're satisfied with an AI response, simply click the New card button to save it as a card

When you want to start researching a new topic, this approach allows you to quickly upload relevant materials and have AI help you plan research directions or summarize content. You can then continue creating your own cards, importing more materials, and engaging in ongoing dialogue with AI to build your own deep understanding.

Chatting with Whiteboards​

Whiteboard with AI chat panel open

In Heptabase, each whiteboard represents a topic you care about. When you open a whiteboard, you can click Chat in the upper right corner to view chats related to this whiteboard. In each chat, you can not only discuss with other collaborators on this whiteboard, but also enable AI mode to converse with the latest AI models from tech companies like OpenAI, Google, and Anthropic.

Whiteboards serve three main purposes for AI conversations:

  1. Thinking tool: When you want to better organize messages from AI, you can directly drag messages you find valuable onto the whiteboard for visualization. As accumulated chats and messages grow, you can use these messages as a foundation to build deeper, more comprehensive understanding of the topic on the whiteboard.

  2. Topic-based folder: You can centralize AI chats related to a specific topic on the same whiteboard for easy future reference.

  3. Knowledge context: You can add the entire whiteboard to the context of your AI chats, allowing AI to answer your questions based on the whiteboard's content. For example, you can add all journals from the past six months to a whiteboard with one click, ask AI to read them and then ask about things you don't know about yourself, or ask for advice based on these journals; or put multiple PDF files and YouTube videos on a whiteboard, letting AI answer your questions based on them.

Specifically, you can follow these steps to start an AI chat:

  • Open a topic whiteboard, click Chat in the upper right corner, where the default mode is AI Chat.

  • Select an AI model and context in Chat, enter your question and start the conversation.

  • Drag messages you find valuable onto the whiteboard and develop more understanding of the topic based on this information.

Providing Context to AI: Add Button and "@" Mention​

Adding context is one of the features that allows AI to deliver maximum value. You can provide context to AI in the following ways:

  • Add button: Click "+" in the left side of the chat box to add specific cards or an entire whiteboard to the context.

    Add button menu showing options to add context

  • @ Mention: Type "@" in a message, search for and select the card, section, or whiteboard you want as context to make it a key reference. This method is suitable for quickly finding specific materials.

    @ mention showing sections list

Chatting with PDFs​

If your learning and research sources are often PDF files (for example: academic workers doing literature reviews, product managers reading industry reports, students reading reference books), through Heptabase's PDF Parser and AI, you can have AI help you understand PDF content, even discuss specific pages and paragraphs with AI, and break key points into cards to place on the whiteboard while maintaining traceability to page numbers and paragraphs.

For example, when you want to discuss a PDF with AI, you can:

  1. Drag the PDF directly onto the whiteboard.

  2. Hover your cursor over the PDF and click Ask AI, or right-click on the PDF and select New Chat to start a conversation with AI.

PDF right-click menu with New Chat option

When chatting for the first time, Heptabase will automatically parse all text, tables, images, formulas, and other content in the PDF into a format suitable for AI to read, so some waiting time is required.

Once parsing is complete, you can specify page ranges when chat with AI based on this PDF. The specific method is as follows:

  1. Confirm that the PDF has been added to the current AI chat context (currently not applicable to PDFs mentioned using @).

  2. Click the PDF title above the chat box.

  3. A range selection menu will appear; select the specified page range.

PDF page range selector

Besides, you can directly select the paragraph you want as context and provide it to AI. In the image below, I selected part of section 3.2, then added the content of that range to the conversation context through the button on the left.

PDF paragraph selection and add to context

If you want to learn more about actual use cases for deep learning in Heptabase, you can read The Best Way to Use AI for Learning; if you want to know how to read and learn from different types of sources in Heptabase, please refer to Read PDFs, media, eBooks & webpages.

Chatting with Videos​

YouTube video with AI chat panel

If you prefer to learn or research through videos (for example: watching videos on specific topics and taking notes, attending online courses, reviewing replays of seminars and lectures), you can discuss specific videos with AI directly on the whiteboard. If this is a YouTube video, please follow these steps:

  1. Copy the YouTube video URL you want to add.

  2. Return to the whiteboard and paste the video using the shortcut (Cmd/Ctrl + V).

  3. Right-click on the video and select New Chat to start chatting with AI.

This way, AI can respond based on the video content and attach timestamps to help you trace back to key segments at any time. Through this method, you can quickly transform key points scattered throughout a video into a digested knowledge structure.

Additionally, if the video you want to discuss with AI is not from YouTube, or if you have other types of learning materials like audio files or podcasts, you currently need to drag the file onto the whiteboard first and use Heptabase's speech-to-text feature to generate subtitles before you can chat with AI based on these files. For these operations, please read the Podcast section of Read PDFs, media, eBooks & webpages.

Chatting with Journals​

Chatting with Journals

As you accumulate more and more content in your Heptabase Journal, you can discuss with AI based on journals from any date range. This is suitable for helping you identify recurring themes, improvement suggestions, and more, digging out unexpected insights from daily records!

  1. Right-click in a blank area of the whiteboard and select Calendar, specifying to add all journals from a specific date range to the whiteboard.

  2. Open Chat in the upper right corner of the whiteboard, click the add button and enter your question, allowing AI to read the entire whiteboard and respond based on your question.

  3. AI's answers will directly locate specific content in the journals.

Chatting with Cards​

Chatting with Cards

Of course, you can also discuss specific cards and paragraphs in Heptabase with AI. Here are several ways to chat with cards:

  1. Select a card on the whiteboard, right-click and select New Chat.

  2. Select text within a card and choose Add to Chat.

  3. Mention specific cards using @ in your messages to AI, having AI answer your questions based on these cards' content.

When using Heptabase's AI chat feature, one of my favorite uses is to edit AI's reply messages, highlighting paragraphs I find important, adding some of my own annotations, and deleting information that's not useful to me. You can click the Edit button in the upper right corner of an AI's reply message to modify its content.

Chatting with Your Entire Space​

Heptabase has a very powerful feature called "Chat with your entire Space." This option is enabled by default, which means that whenever you ask AI a question, it will search for relevant data across your entire Space and attempt to answer your question.

You'll see AI's thought process showing which content it's trying to search and read. For example, when you ask it to analyze what you've been researching over the past month, it might reference your journal cards from the past month. When you ask about specific topic keywords, it will search for corresponding cards, whiteboards, and PDFs. If you want more precise control over what data AI references, you can temporarily turn off the Space option and manually add the data you want it to reference.

To learn more about how AI accesses your data when "Space" is enabled, please refer to What data can AI access when I turn on the "Space" option in the AI chat?

AI Credits​

Currently, as long as you're a Heptabase subscriber, you receive some free AI credits from Heptabase each month to use for chatting with AI. Therefore, if you've read this far, we recommend you immediately try using the Research a topic feature to import some materials you want to research and learn about, and discuss them with AI.

If you're subscribed to the Pro plan, the monthly free credits are limited. If you want more AI credits, you can consider upgrading to the Premium plan to enjoy the following benefits:

  • Unlimited usage of basic models like Gemini 3 Flash and GPT 5-mini when you run out of AI credits.
  • 10× the AI usage limit of the Pro plan
  • Access to all the latest models from Google, OpenAI, and Anthropic (the Pro plan only provides access to Gemini series models)

For detailed information about pricing and plans, please refer to Pro and Premium Plans: Pricing & FAQ.

Advanced Features: Model Context Protocol (MCP)​

For advanced users, Heptabase supports the Model Context Protocol (MCP), which allows external AI services (such as ChatGPT or Claude) to directly read, search, and write to your Heptabase knowledge base. This creates a seamless loop between thinking with AI and growing your long-term knowledge in Heptabase.

To learn more about setting up and using MCP, please refer to: How to use Heptabase MCP.