Skip to main content

Work with AI

Untitled

TL;DR​

In Heptabase, you can chat directly with AI, manually edit and annotate important AI responses, then drag them onto whiteboards for visual organization. Additionally, you can use the plus button at the top left of a message or the "@" feature to have the AI answer based on the content of your whiteboards, notes, or video contents, enabling systematic and deeper research into topics.

Introduction​

In the age of AI, people commonly use well-known AI apps like ChatGPT, Google Gemini, and Claude to learn or research topics they care about. However:

  1. When exploring complex topics involving multiple interconnected concepts, you might want to extract knowledge provided by AI from linear chats and place it on a large whiteboard for analysis and future review.

  2. When asking questions, you often want AI to reference your previously written notes or saved knowledge content.

These scenarios are precisely what Heptabase's AI features excel at.

Integrating Whiteboards and AI​

In Heptabase, each whiteboard represents a topic you're interested in. When you open a whiteboard, you can click the Chat button at the top-right corner to view all chats belonging to or mentioning that whiteboard. In each chat, you can chat not only with collaborators on the whiteboard but also enable AI mode to interact with the latest AI models from companies like OpenAI, Google, and Anthropic.

Untitled

Whiteboards serve three main purposes in AI chats:

  1. Thinking Tool: To better organize messages received from AI, you can directly drag valuable messages onto the whiteboard for visualization. As your chat messages and dragged content grow, you'll be able to build a deeper and more comprehensive understanding of the topic on your whiteboard.

  2. Topic-based Folder: You can organize all AI chats related to a particular topic within the same whiteboard, making future reference easier.

  3. Knowledge Context: You can use the plus icon at the top left of the chat window to add an entire whiteboard to the AI chat context. This allows the AI to answer your questions based on the content of the whiteboard. For example, you can throw all the journals you’ve written in the past six months onto the whiteboard with one click, let the AI read them, and then ask it things about yourself that you don’t know, or get advice based on these journals.

Untitled

Integrating Cards and AI​

In addition to simply chatting with AI, you can also use “@” to mention your cards in the message you send to AI. The AI will then use the content of those mentioned cards as the basis for answering your question. Currently, the card types that AI can read include note cards, PDF cards, video cards with subtitles, journal cards, highlight cards, and more. For note cards and PDF cards, you can even select specific text ranges or page numbers when asking the AI a question.

Untitled

Heptabase’s AI features are currently in public beta, and users need to input their own API key to use them. In the coming months, we’ll gradually provide free AI usage credits to Pro Plan users, as well as more advanced AI plans for those who prefer not to use their own API key.

At the same time, we’ll also introduce a number of advanced AI features, such as:

  • Allowing AI to answer questions based on your entire knowledge base or other MCP services;

  • Enabling AI to edit and manipulate cards and whiteboards according to your instructions;

  • Allowing other AI services (e.g., ChatGPT) to read and write Heptabase data through the Heptabase MCP Server;

  • …

Stay tuned!