Skip to content

Add library-cli: YouTube knowledge library manager#12

Open
DTrejo wants to merge 3 commits into
mainfrom
library-cli
Open

Add library-cli: YouTube knowledge library manager#12
DTrejo wants to merge 3 commits into
mainfrom
library-cli

Conversation

@DTrejo
Copy link
Copy Markdown
Owner

@DTrejo DTrejo commented Nov 25, 2025

note to readers: I barely scanned this code and didn't do any refactoring for styling or organization. so this is a pretty good representation of how js / deno code comes out without any refactoring prompting (essentially just vibe coded).

Adds a Deno CLI tool for building a searchable knowledge library from YouTube videos.

Features

  • Downloads video metadata, descriptions, chapters, and transcripts (no video files)
  • Uses yt-dlp for YouTube data extraction
  • JSON3 transcript format (eliminates duplication)
  • Full-text search with ripgrep
  • Color-coded, Google-like search results
  • Organized by date with dasherized folder names

Commands

  • library add [url] - Add video to library
  • library search [query] - Search with ripgrep
  • library --help - Show help

Installation

cd library-cli
deno install --global --config deno.jsonc --allow-read --allow-write --allow-env --allow-run --allow-net --name library --force main.ts

Includes 6 passing smoke tests.

Development Cost Analysis

Raw Token Usage:

  • Input: 1.1K tokens
  • Cache Creation: 260.8K tokens
  • Cache Read: 24.8M tokens
  • Output: 75.8K tokens
  • Total: 4,283,040 tokens (session: 142,622 effective tokens)

Session: 34m 55s (time spent waiting for single agent)

At 6 hours/day of coding (pricing):

  • $20/mo (20M tokens): ~13.6 days (~0.5 months) — $1.47/day
  • $200/mo (200M tokens): ~136 days (~4.5 months) — $1.47/day

Note: Caching dramatically reduces effective token usage. Sessions like this use ~10/day at 6hr/day pace.

Videos in Library

Summary of Videos: ai engineering todo list extracted from the video descriptions

DTrejo and others added 3 commits November 24, 2025 20:00
- Deno CLI tool for building searchable video library
- Downloads YouTube metadata, descriptions, chapters, transcripts (no video)
- Uses yt-dlp for metadata extraction, ripgrep for search
- JSON3 format for transcripts (no duplication)
- Color-coded output, Google-like search results
- Organized by date with dasherized folders
- Includes 6 passing smoke tests

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant