@eventcatalog/generator-ai NPM

Read the Docs | Edit the Docs | View Demo

Core Features

🗣️ Use natural language to ask questions about your EventCatalog
⭐ Choose your embedding model
📚 Reduce the time it takes to learn and understand your architecture
⭐ And much more...

How it works

EventCatalog is technology agnostic, meaning it can integrate with any schemas, specs or brokers.

EventCatalog supports generators. Generators are scripts are run to pre build to generate content in your catalog. Generators can use the EventCatalog SDK.

With this EventCatalog plugin embeddings are generated based on your EventCatalog. You can either use huggingface or openai to generate the embeddings.

These embeddings are stored and use by EventCatalog, all locally in your project and on your computer. Once you generate the embeddings with this generator, and configure your EventCatalog project, you can use the Chat to ask questions about your catalog.

EventCatalog integrates with OpenAI Models or you can use our browser based models.

These models will use the embeddings which are generated by this generator.

You can pick which model and embedding provider you want to use, and EventCatalog will download the model for you enabling you to use the Chat.

...
generators: [
    [
      '@eventcatalog/generator-ai',
      {
        // This will split the markdown files into smaller chunks, this is optional and defaults to false
        // This uses MarkdownTextSplitter under the hood
        splitMarkdownFiles: false,


        // Set this to true if you want to include users and teams in the embeddings for your AI Assistant
        // This can bloat the embeddings and may effect the results you get back, false by default
        includeUsersAndTeams: false,

        // Should the schemas for the messages also be embedded?
        // Default value is true. (You must specify the schemaPath in your message for the schema to be embedded)
        includeSchemas: true,

        embedding: {
          // defaults to huggingface
          provider: 'openai' | 'huggingface',
          //  pick your model (optional, defaults to text-embedding-3-large for openai and all-MiniLM-L6-v2 for huggingface)
          model: 'text-embedding-3-large' | 'Xenova/all-MiniLM-L6-v2',
        },

        // optional similarity results to use, defaults to 30
        similarityResults: 40,

        // optional max tokens to use, defaults to 4096, map to your model
        maxTokens: 4096,

      },
    ]
  ],
...

In this example we install and setup the generator. This will parse all our catalogs files and create a new folder called /generated-ai with the following files:

documents.json - Contains the processed documents and their metadata for the LLM model in EventCatalog
embeddings.json - Contains vector embeddings for the documents. These are used by EventCatalog to do vector searches.
README.md - This file just tells you that these files are generated, and not to edit them

Getting started

Installation and configuration

Make sure you are on the latest version of EventCatalog.

Install the package

@eventcatalog/generator-ai

Configure your eventcatalog.config.js file (see example)
Run the generate command

npm run generate

Configure your EventCatalog project, turn on the chat feature.

{
    chat: {
        enabled: true,
    }
}

Start your project and go to /chat to see the AI Assistant or use the link on the sidebar.

npm run dev

Found a problem?

Raise a GitHub issue on this project, or contact us on our Discord server.

Gold sponsors

Sponsors help make EventCatalog sustainable, want to help the project? Get in touch! Or visit our sponsor page.

Enterprise support

Interested in collaborating with us? Our offerings include dedicated support, priority assistance, feature development, custom integrations, and more.

Find more details on our services page.

Contributing

If you have any questions, features or issues please raise any issue or pull requests you like. We will try my best to get back to you.

You can find the contributing guidelines here.

Running the project locally

Clone the repo
Install required dependencies pnpm i
Run tests pnpm run tests

Commercial Use

This project is governed by a dual-license. To ensure the sustainability of the project, you can freely make use of this software if your projects are Open Source. Otherwise for proprietary systems you must obtain a commercial license.

If you would like to obtain a Commercial License, you can purchase a license at https://eventcatalog.cloud or email us at hello@eventcatalog.dev