Summarize text and public webpages using OpenAI API.
When configuring the plugin, we recommend using the LLM model with a context window size of 16k tokens or more. This will let you summarize webpages and long texts.
At the moment of writing, the most balanced model is gpt-3.5-turbo-1106
in terms of price and quality.
The full list of models can be found here.
This repository contains the plugin's source code.
Path | Description |
---|---|
./src/index.ts | The entry point for the plugin. It contains the plugin definition and references to all the actions. |
./src/actions/ | This folder contains all the actions of the plugin. Each action is represented by a separate file with the action definition and implementation. |
This plugin is built using Connery SDK, the open-source SDK for AI plugins and actions development.
Learn how to use the plugin and its actions.
If you have any questions or need help with this plugin, please create an issue in this repository.