doXmind

doXmind is an AI-native writing editor that thinks with you. Write, research, and present — with built-in AI chat, knowledge base, database blocks, and presentation mode.

Comment

Inspiration

As students who write research papers, reports, and documentation daily, we were frustrated with existing writing tools. Grammar checkers only fix surface-level errors. AI chatbots require constant copy-pasting between apps. Traditional editors like Notion and Google Docs treat AI as an afterthought — a bolt-on feature, not a core part of the experience.

We dreamed of an intelligent writing companion that could truly understand our content and help us think — not just correct typos. That's when we decided to build doXmind: an AI-native editor designed from the ground up.

What it does

doXmind is an AI-native writing editor with intelligence built into every layer:

  • AI Chat: Discuss your document with doxmind AI — it reads your full content, suggests improvements, and makes edits directly in the editor
  • AI Thinking Mode: Watch the AI reason step-by-step with visible chain-of-thought before answering complex questions
  • Quick Edit: One-click commands to fix grammar, improve clarity, simplify, expand, or translate selected text
  • Smart Autocomplete: Context-aware writing suggestions as you type, like GitHub Copilot for prose
  • Inline Diff Review: Review every AI change with accept/reject controls — you always stay in control
  • Knowledge Base: Upload PDFs and documents as reference materials; doxmind AI uses them to ground its responses with source citations
  • KB Agent: Ask questions across all your uploaded documents — AI synthesizes answers with relevance scores and source references
  • Database Blocks: Notion-style structured data with table, board, gallery, and list views — right inside your documents
  • CSV Import & AI Analysis: Drop a CSV file and get instant AI-powered charts and data insights
  • Inline Comments: Threaded discussions anchored to specific text, with resolve/unresolve support
  • Presentation Mode: Turn any document into a full-screen slideshow with one click
  • Semantic Search: Search by meaning, not just keywords — find related content across all your documents
  • AI Writing Review: Real-time feedback on grammar, clarity, tone, and engagement

How we built it

Frontend: React 19 + Vite + TailwindCSS + Framer Motion for a fast, responsive UI. We chose TipTap as our rich-text editor for its extensibility — we built custom blocks for databases, multi-column layouts, web bookmarks, and more. Recharts powers data visualization. i18next enables full internationalization across 6 languages.

Backend: FastAPI handles API requests with Server-Sent Events (SSE) for real-time AI streaming responses. JWT authentication with automatic token refresh keeps sessions secure.

AI Integration: doxmind AI powers all features — chat, autocomplete, quick edits, thinking mode, and document review. We implemented RAG (Retrieval-Augmented Generation) with vector embeddings for the knowledge base, including an agentic KB search that provides source citations and relevance scores.

Database: PostgreSQL for document storage, user data, database blocks, and knowledge base indexing.

Infrastructure: Regional API routing (global + China servers), structured SEO with JSON-LD schemas, and responsive design for mobile and desktop.

Challenges we ran into

  1. Streaming UX: Implementing smooth real-time AI streaming while maintaining editor state was our biggest technical challenge. We solved this with careful state management using SSE and TipTap's transaction system.

  2. Context Management: Fitting large documents + knowledge base + conversation history within token limits required smart chunking and prioritization strategies.

  3. Diff Review System: Building an intuitive inline diff view for reviewing AI suggestions took multiple iterations. Users need to see exactly what changed and easily accept or reject individual edits.

  4. Database Blocks in a Rich-Text Editor: Implementing Notion-style database views (table, board, gallery, list) inside TipTap required custom node extensions and complex state management.

  5. Internationalization: Supporting 6 languages (English, Chinese, French, Japanese, Korean, Spanish) across the entire app — including dynamic AI-generated content — required a comprehensive i18n architecture with i18next.

What we learned

  • Extended context windows are a game-changer for document-aware AI — enabling cross-document knowledge base search that wasn't practical before
  • Users want control over AI suggestions, not blind acceptance — our inline diff review became one of the most-loved features
  • AI thinking mode (showing reasoning steps) dramatically increases user trust in AI suggestions
  • The best AI tools feel like collaborators, not replacements — AI should think with you, not for you
  • Building an AI-native editor from scratch enables much deeper integration than bolting AI onto an existing tool

What's next for doXmind

  • Real-time collaboration — multiple users editing the same document simultaneously
  • More export formats — LaTeX, EPUB, and advanced PDF layouts
  • Custom AI personas — different writing styles and expertise areas for specialized domains
  • Mobile apps — native iOS and Android editors
  • API & integrations — connect doXmind with other tools in your workflow
  • Advanced data analysis — more chart types, pivot tables, and AI-powered data stories

Built With

Share this project:

Updates