YouTube Search and Summarize

This is a project of mine that I am now using to showcase the usage of Promptate for generating ads in a YouTube search and summarize OpenAI Plugin. It provides an API to search for YouTube videos and retrieve their transcripts. It fetches ads from Promptate for the user.

Additionally, I have also integrated another one of my projects ProdigiMetrix into it. It tracks basic metrics like time and keywords when routes are used. If you want this functionality in yours, follow the credentials section in the readme to get your own account and start capturing valuable user insights.

Overview

The YouTube search and summarize plugin is built using Python and Quart framework for asynchronous web serving. The service allows users to perform the following actions:

1. Search for YouTube Videos: Users can perform a search query to find videos on YouTube.

2. Get Video Transcripts: Users can obtain the raw transcript of a specific YouTube video by providing its URL.

3. Get Next Page of Search Results: Users can retrieve the next page of search results for a particular search query.

4. Fetch Ads from Promptate: The service integrates with Promptate to fetch relevant ads for the users.

5. Track Keywords with ProdigiMetrix: You can track prompt times, keywords, and overall activity using the ProdigiMetrix Service.

Getting Started

To run the service on your local machine, follow these steps:

1. Install Dependencies: First, install the required Python packages using the following command:

   pip install -r requirements.txt
   

2. Start the Service: Run the Python script app.py to start the service:

   python app.py
   

   The service will be available at http://localhost:5000.

API Endpoints

The service provides several API endpoints that can be accessed to interact with the YouTube search and summarize functionality. Below are the available endpoints:

Get Video Transcript

• Endpoint: POST /search/youtube/transcript
• Input: JSON body with url field containing the YouTube video URL.
• Output: JSON response containing the transcript of the video.

Search YouTube Videos

• Endpoint: POST /search/youtube/query
• Input: JSON body with query, keywords, iso_3166-1_alpha-2_code, and iso_639-1_code fields.
• Output: JSON response containing videos list and a query_id to get the next page of results.

Get Next Page of Results

• Endpoint: POST /search/youtube/next
• Input: JSON body with query_id obtained from the /search/youtube/query endpoint.
• Output: JSON response containing the videos for the next page of results.

Miscellaneous Endpoints

• /search/youtube/logo.png: Get the app logo.
• /.well-known/ai-plugin.json: OpenAI plugin manifest.
• /search/youtube/openapi.yaml: OpenAPI specification.
• /search/youtube/ping: A simple health check.

Ad Integration with Promptate

The service includes a function get_promptate_ad(user_id) that fetches ads from Promptate based on the provided user_id. Currently, it uses hardcoded values for plugin_name and developer_token, but in a real-world scenario, you should register with Promptate to obtain your own tokens and handle them securely.

Customization

You can customize the service according to your requirements. For instance, you may want to implement user authentication, handle more search parameters, or store user preferences in a database. The service is provided as a basic template, and you can extend it as per your needs.

You can also customize the way the keywords are being tracked with ProdigiMetrix. Right now keywords are submitted, but maybe your plugin doesn't need that. Maybe you want to track which feaure is being used. To do this, just submit a single keyword like search, next, or transcript, to find out what users are using.

Please note that this project serves as a demonstration of Promptate usage and does not represent a full-fledged production-ready service. It is intended to provide developers with a starting point for building their own plugins and services using Promptate and YouTube APIs.

License

This project is licensed under the MIT License. Feel free to modify and use it as per your needs.

Acknowledgments

This project uses the following third-party libraries:

• youtube-transcript-api: A Python library to fetch YouTube video transcripts.
• youtube-search-python: A Python library to perform YouTube video searches.
• requests: A Python library to make HTTP requests.
• Quart: A Python asynchronous web framework.
• Quart-CORS: A Quart extension for handling CORS.
• tiktoken: A Python library to count the number of tokens in a text string.
• prodigi_metrix: My metrics tracking package for AI projects.

Please make sure to give proper credit to the authors and contributors of these libraries and include any additional licensing information if required.

Note

The integration with Promptate is for demonstration purposes and uses hardcoded values. In a real-world application, you should follow the recommended practices for securing tokens and integrating with external services.

The integration with ProdigiMetrix is for demonstration purposes and uses .env variables meant for this project only. If you need this functionality, follow the instructions in the credentials section of the readme.