Skip to content

Search in real-time for snippets from YouTube transcripts.

License

Notifications You must be signed in to change notification settings

0xRy4n/transcripter

Repository files navigation

Transcripter

Linting and Documentation Codacy Badge Publish to PyPI License: MIT

Transcripter is a tool for indexing and searching YouTube video transcripts. It makes use of the YouTube Data API to fetch video details and transcripts, and collates and stores the transcript chunks in Redis for efficient full-text search.

Transcripter's main purpose is to provide a simple way to add transcript-search functionality to your application. Inspired by (but unrelated to) https://ippsec.rocks/.

Structure

  • transcripter/: Main package
    • core/: Core functionality
      • redis_manager.py: Manages Redis operations
      • youtube_manager.py: Handles YouTube API interactions
    • services/: Service layer
      • indexing_service.py: Manages the indexing process
      • search_service.py: Manages the search process
    • config.py: Configuration management
    • logs.py: Logging management
  • examples/: Example applications
    • basic/: Basic example
      • app.py: Simple application demonstrating usage
  • index-config.yml: Configuration file for indexing. Used by the example app.
  • docker-compose.yml: Docker compose file for the example app.

How It Works

  1. The IndexingService periodically fetches video details and transcripts from configured YouTube playlists, channels, or individual videos.
  2. Transcripts are processed and stored in Redis using the RedisManager.
  3. The YouTubeManager handles interactions with the YouTube API.
  4. Indexed data can be searched efficiently using Redis's full-text search capabilities.

Installation

You can install from PyPI:

pip install transcripter

Manual Installation

make install

or

pip3 install git+https://github.com/0xRy4n/transcripter

Note: Transcripter requires a Redis Stack server to run. It's recommended to use the official Docker image like so:

docker run -d --name redis-stack-server -p 6379:6379 redis/redis-stack-server:latest`

You can run the Redis Stack either locally or remote, depending on your preference.

Configuration

Environment Variables

Set the following environment variables:

  • YOUTUBE_API_KEY: Your YouTube API key
  • REDIS_HOST: Redis server host
  • REDIS_PORT: Redis server port
  • REDIS_PASSWORD: Redis server password (if applicable)
  • REDIS_INDEX: Name of the Redis index to use
  • TRANSCRIPTER_LOG_LEVEL: Log level (DEBUG, INFO, WARNING, ERROR, CRITICAL)

Indexing Configuration

Configure the indexing process in your application by specifying a indexing-config.yml file like so:

# Change me
sources:
  playlists:
    # playlist id
    - PLlkHl5i0GaAlBvAhi9R7S4vQ0IhR3ePCU
  channels:
    - # channel id
  videos:
    - # video id

indexing:
  interval: 3600  # Reindex every hour

You can then pass the path to this file in to the Config object:

config = Config(config_path='path/to/indexing-config.yml')

By default, it will search for a file called indexing-config.yml in the root directory of the repository (one directory up from the transcripter directory).

Alternatively, you can simply instantiate Config without any arguments, and set each property directly:

from transcripter.config import Config

config = Config()
config.playlists = ['playlist_id1', 'playlist_id2']
config.channels = ['channel_id1', 'channel_id2']
config.videos = ['video_id1', 'video_id2']
config.indexing_interval = 3600  # Indexing interval in seconds

Starting the Indexing Service

from transcripter.services.indexing_service import IndexingService

indexing_service = IndexingService(config)

Indexing your videos is easy.

indexing_service.index_all()

Once a video has been indexed, it will not be indexed again, and it's transcript won't be pulled on successive runs.

You'll likely want to run the indexing service periodically as a background job. It is up to you to decide how you would like to implement this.

Searching

You can search the indexed data using the SearchService class.

from transcripter.services.search_service import SearchService

search_service = SearchService(config)

You can then search for videos by title:

search_service.search('search query')

It will return a list of stored transcript chunks in the following format:

[
  {
    "snippet": "directory don't worry about it too much it's not going to be on the test there's not going to be a test but if there were",
    "start_time": 300.56,
    "timecode": "00:05:00",
    "video_id": "_haJQY-Y70E",
    "video_title": "Linux Structure and Commands - Intro to Hacking w/ HTB Academy #4"
  },
  {
    "snippet": "directory don't worry about it too much it's not going to be on the test there's not going to be a test but if there were",
    "start_time": 300.56,
    "timecode": "00:05:00",
    "video_id": "_haJQY-Y70E",
    "video_title": "Linux Structure and Commands - Intro to Hacking w/ HTB Academy #4"
  }
]

Example Web App

A fully functional example web app is included in the examples/basic directory. It is a simple web app that allows you to search for videos by title. It is built with Flask and uses Bootstrap for the frontend.

You can change the index-config.yml file in the root of the repository to configure the playlists, channels, and videos you want to index.

Additionally, you can opt to either create a .env file in the root of the repository and set the environment variables directly, or set them in your operating system's environment variables.

The example app is containerized with Docker, making it simple to run:

docker compose up

Alternatively, you can run the app without Docker by installing the dependencies and running the app.py directly:

First you'll need to run Redis Stack in a docker container:

docker run -d --name redis-stack-server -p 6379:6379 redis/redis-stack-server:latest

Then install the dependencies and run the app:

pip3 install -e .
python3 examples/basic/app.py