Daftar

Logo
Search the literature directly from your note, with results shown alongside your writing for a seamless research workflow.
Daftar generates AI-powered query ideas from your note, helping turn rough thoughts into focused research directions.
A slash-command editor makes it easy to insert structured blocks like headings, lists, and quotes while drafting notes.
AI relevance explanations show why a paper matters to your note, helping users judge sources faster and stay grounded.
Daftar brings note-taking, organization, and literature discovery into one desktop-style workspace built for active research.
Write technical notes with rich formatting and clean code blocks, making Daftar useful for both concepts and implementation details.

Inspiration

We wanted to make research feel less scattered. Most note apps help you write, and most research tools help you find papers, but very few do both together. We built Daftar to connect writing and discovery.

As you write, your notes become useful context. Instead of opening new tabs, rewriting searches, and manually saving sources, Daftar helps turn freewriting into research.

We’re all South Asian, and that is where the name comes from. “Daftar” is a common word across South Asian languages that can mean notebook, record, or office. It fit a project built around writing, organization, and ideas.

What it does

Daftar is an AI-powered research notebook.

It lets users:

write and organize notes
use rich text or LaTeX
generate query ideas from a note with Gemini
search papers with Semantic Scholar
get paper recommendations based on the open note
save references to each note
view note history

The main idea is simple: your note is not just text, it helps drive discovery.

How we built it

We built Daftar with React, Vite, TypeScript, Tailwind CSS, Zustand, Electron, Node.js, Gemini, Semantic Scholar, TipTap, CodeMirror, and KaTeX.

The frontend handles the note-taking interface and app state. TipTap powers rich text editing, CodeMirror supports LaTeX, and KaTeX renders math.

The backend is a lightweight Node server that connects to Gemini and Semantic Scholar. Gemini helps generate queries and explain relevance. Semantic Scholar powers paper search and recommendations. Electron wraps everything into a desktop-style app.

We also built separate literature flows for:

manual search
note-based recommendations
similar papers
saved references

Challenges we ran into

One challenge was debugging the frontend, backend, and Electron setup. Sometimes the UI worked while the backend was missing API keys, which made issues harder to trace.

Another challenge was making the literature experience clear. Search, recommendations, query generation, and similar papers are related, but they are not the same. We had to make those flows feel distinct without being confusing.

We also had to keep the project maintainable while moving fast. Some important files grew large, especially in the literature sidebar, so we cleaned them up without changing behavior.

Accomplishments that we're proud of

We’re proud that Daftar already feels like a real product, not just a rough prototype.

Some highlights:

making AI feel actually useful
turning note content into paper recommendations
supporting both rich text and LaTeX
creating a desktop-style research workflow
combining search, recommendations, and saved references in one place
cleaning up the codebase enough to keep building on it

What we learned

We learned that the hard part of AI products is not just using the model. It is building an experience where users understand what the AI is doing and why.

We also learned about: