Editor Assistant for Confluence

Bring the creative power of AI to your Confluence instance. Conveniently edit any page with auto-complete suggestions from ChatGPT, quick-fix auto-suggestions, or even any question you want.

Comment

Inspiration

Inspired by the power of code completion tools like GitHub Copilot, we envisioned a transformative experience for text editors. The idea of a text editor intuitively completing articles, much like code editors do with code, piqued our interest. Additionally, integrating a feature that allows users to interact with ChatGPT directly by selecting text sections seemed like a groundbreaking way to enhance user engagement and improve content quality.

What it does

Editor Assistant for Confluence offers:

  • Sentence Completion: Seamlessly auto-completes your sentences, making content creation faster.
  • Quick Fix: Improves your text based on your preference - be it fixing grammar, adopting a more formal tone, or even making it concise.
  • AI Assistant: A unique feature that allows users to ask AI-driven questions about their article or a particular selection. The AI can also suggest modifications, and upon approval, can automatically implement them.

How we built it

  • Platform & Integration: Leveraging Atlassian's Forge platform, we integrated the ChatGPT API. This architecture enables direct communication between Forge and ChatGPT, negating the need for any intermediary infrastructure.
  • Tech Stack: We employed a robust stack consisting of TypeScript and React to bring our vision to life.

Challenges we ran into

  • Time Constraints: The Forge platform has a stringent timeout limit for backend function invocations. This posed a challenge when awaiting lengthy responses from ChatGPT.
  • Streaming Limitation: Forge's restrictions prevented us from utilizing the streaming feature offered by the ChatGPT API. This constraint impacted the user experience, especially during real-time interactions.
  • Editor Reimplementation: The absence of a plugin API in Confluence that could enhance the built-in editor meant that we had to recreate the Confluence editor using the atlaskit package.

Accomplishments that we're proud of

  • Secure Integration: We ensured end-to-end security by enabling users to input their ChatGPT keys directly, ensuring no third-party platforms, other than Atlassian and OpenAI, handle the keys.
  • Intuitive UI: Our graphical user interface stands out with its user-friendly auto-completion panel and a context-aware quick-fix button.

What we learned

  • AI as a Catalyst: AI tools, including code generation with ChatGPT and logo design with DALL·E 3, are game-changers. They significantly accelerated our development process, proving indispensable for rapid project completion.
  • Editor Mechanics: Delving deep into the atlaskit/editor package and the prosemirror editor enriched our understanding of the intricacies of text editor operations.

What's next for Editor Assistant for Confluence

Our roadmap includes addressing the current limitations. We're considering transitioning to our infrastructure, which will alleviate the timeout issue and enhance the user experience. This shift would also allow us to leverage features like streaming tokens directly from the backend, bypassing platform constraints.

Share this project:

Updates