README.md

yt-chat is a tool designed to help you summarize any Youtube video.

Once a video is summarized, you can also ask more precise questions about the video in question.

---

Try yt-chat!

You can play with yt-chat right away by visiting yt.chat. You can summarize up to 5 videos per day.

Installation

This section is useful if you want to install yt-chat on your machine.

After cloning the repository, and with poetry installed, run the following command from the repository root:

poetry install

To run yt-chat, simply do:

poetry run chainlit run yt_chat/app.py -w

If you don't want to bother, you can also use try the online version (only handles OpenAI models, i.e. GPT-3.5 and GPT-4).

Using ChatGPT-3.5

If you wish to use an OpenAI model, for example gpt-3.5, you will need your OpenAI API key.

Once you've input your OpenAI API key requested by yt-chat, select the ChatGPT chat profile in the UI.

Using Mistral-7B

If you wish to use a local ollama model, for example mistral-7b, you will need to install ollama on your machine.

First, make sure your ollama server is running. Then, run yt-chat (when running yt-chat for the first time, you will asked for an OpenAI API key; this is irrelevant for local models, enter anything to continue).

Once yt-chat is running, simply select the Mistral chat profile in the UI.

Configuration

Check out yt_chat/config.py and yt_chat/config_prompts.py for configuring the app parameters and prompts.

Docker

We provide Docker support:

docker-compose up -d --build

Acknowledgments

yt-chat is powered by chainlit, qdrant, and ollama.