Skip to content

dnyanadapathare/vaultdrop

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Vaultdrop UI

⬡ vaultdrop

The missing capture layer for your Obsidian second brain.

Paste any YouTube, Instagram, Substack, or Reddit URL — get a structured markdown note in your Obsidian vault automatically.



The problem

Most knowledge systems fail at capture. You save a reel, a thread, an article — and it sits in your browser bookmarks, Instagram saved posts, or YouTube watch later. It never makes it into Obsidian. It never becomes a note you can think with.

Capture friction kills PKM habits. If getting something into your vault takes more than 10 seconds, you stop doing it.

Vaultdrop solves the capture layer for video and social content — the piece no other tool handles.


What it does

Paste a URL. Get a structured .md note in your Obsidian vault — with frontmatter, category, tags, TL;DR, key insights, and full notes. Automatically.

Platform How it works
YouTube Shorts Extracts captions → falls back to audio transcription
Instagram Reels Downloads audio → transcribes with Sarvam AI
Substack articles Scrapes full content → structures with Groq
Reddit threads Pulls post + top comments via JSON API
Any article Generic scraper for blogs and web pages
Manual paste Fallback for anything that can't be extracted

Every note lands in your Inbox/ folder with this structure:

---
title: "Pigeons Diagnose Cancer"
source: "https://youtube.com/shorts/..."
platform: youtube
saved: 2026-05-13
category: "[[Science & Nature]]"
tags: ["cancer", "pigeons", "ai"]
---

## TL;DR
Pigeons were trained to diagnose breast cancer with 85% accuracy...

## Key Insights
- Pigeons can be trained to diagnose breast cancer with high accuracy
- Combining pigeon votes increases accuracy to 99%
- This challenges the idea that humans are uniquely capable of certain tasks

## Full Notes
A 2015 study trained pigeons to identify breast cancer in pathology slides...

Graph view — automatic clustering

Every note links to a category index note via [[wikilinks]]. Save 10 recipe reels and they all cluster around a Recipes & Food node in your graph automatically.

20 default categories including: Recipes & Food, Travel, Health & Fitness, Technology & AI, Career & MBA, Self-help & Mindset, and more. Add your own via CUSTOM_CATEGORIES in .env.


Two capture modes

Mode Best for What it does
Summarise Reels, Shorts, Reddit Key insights + detailed notes
Full capture Articles, Substack, tutorials Complete content preserved

Auto-detects the right mode. Override with the toggle anytime.


Indian language support

Vaultdrop uses Sarvam AI saaras:v3 for audio transcription — built for Hindi, Hinglish, Tamil, Telugu, Kannada, and all Indian languages. A Hindi reel gets transcribed and translated to English automatically. No other PKM capture tool does this.


How I use it

I save around 20-30 pieces of content a week — reels, shorts, threads, articles. Most of it used to disappear into Instagram's saved folder or YouTube watch later, never to be found again.

Now everything lands in my Obsidian Inbox as a structured note. I can search across what I've saved, see connections in graph view, and find that recipe or productivity tip when I actually need it.

The next step I'm building toward: Claude Desktop connected to my vault via MCP. The goal is to ask Claude things like "what have I saved about productivity this month" or "find everything I've captured about travel in Southeast Asia" and get answers grounded in my actual notes — not hallucinated.

Vaultdrop is the capture layer. Claude is the retrieval layer. The vault becomes a thinking partner instead of a graveyard.


Why local-first

  • Your notes are yours — plain .md files, forever
  • Nothing stored on any server
  • Free — you bring your own Groq and Sarvam API keys (both have generous free tiers)
  • Works with Obsidian's full ecosystem — graph view, Dataview, Smart Connections, Claude MCP

Prerequisites

Install once:

  • Node.js v18+
  • yt-dlpwinget install yt-dlp or brew install yt-dlp
  • ffmpegwinget install ffmpeg or brew install ffmpeg
  • Obsidian — create a vault first

Free API keys (no credit card):


Setup

git clone https://github.com/dnyanadapathare/vaultdrop.git
cd vaultdrop
npm run setup

The setup script checks prerequisites, collects your API keys, sets your vault path, and installs everything. Then:

npm run dev

Open http://localhost:5173.


Manual setup

Create a .env file in the root:

GROQ_API_KEY=your_groq_key_here
SARVAM_API_KEY=your_sarvam_key_here
VAULT_PATH=C:\Users\yourname\Documents\ObsidianVault\Inbox
PORT=3001

Mac/Linux: VAULT_PATH=/Users/yourname/Documents/ObsidianVault/Inbox

Add custom categories:

CUSTOM_CATEGORIES=Spirituality,Legal,Architecture

Querying your notes

Dataview plugin:

TABLE title, platform, saved FROM "Inbox"
WHERE contains(tags, "recipe")
SORT saved DESC

Claude Desktop + MCP: Connect Claude to your vault and ask "summarise everything I've saved about productivity this month"

Smart Connections plugin: Semantic search across all notes in natural language.


Project structure

vaultdrop/
├── client/               # React + Vite frontend
│   └── src/
│       ├── App.jsx
│       └── App.css
├── server/               # Express local server
│   ├── index.js          # routes + orchestration
│   ├── groq.js           # Groq summarisation
│   ├── sarvam.js         # Sarvam transcription + chunking
│   ├── vault.js          # writes .md to Obsidian
│   └── extractors/
│       ├── ytdlp.js      # YouTube + Instagram
│       └── scraper.js    # articles + Reddit
├── scripts/
│   └── setup.js          # interactive setup wizard
└── .env                  # your keys (never committed)

Roadmap

  • Browser extension — vaultdrop any page in one click
  • Batch processing — paste multiple URLs at once
  • Claude MCP setup guide
  • Windows/Mac auto-startup
  • Obsidian plugin

Built by

Dnyanada Pathare — product designer, Bangalore.


License

MIT — use it, fork it, build on it.

About

Local content-to-Obsidian pipeline. Paste any URL, get a structured markdown note.

Topics

Resources

Stars

2 stars

Watchers

1 watching

Forks

Packages

 
 
 

Contributors