Quickstart
Lunar is a low-code platform for the construction of AI expert agents and systems. It is designed to enable domain experts to directly build and adapt complex analytical AI-based workflows to approach domain-specific problems. Lunar facilitates the integration of generative AI within complex analytical areas, including but not limited to bioinformatics, engineering, finance, marketing, media, etc.
This quickstart shows how to:
- Setup Lunar on your local environment
- Use some of the Lunar components to create a first workflow
- Interact with a Lunar workflow
- Use a Lunar sub-workflow
- Create your first Lunar custom component
Let's get started ...
Local Lunar setup
The following setup works for Mac, Linux and Windows with WSL2
Windows users can install WSL2 by following the instructions here
Lunar consists of three parts:
- The Lunarcore service - the core of Lunar, a Python-based service that includes the backend functionalities.
- The Lunarverse - Lunar is organized as a plugin-architecture and the Lunarverse represents a collection of components (or plugins) available to create workflows and AI expert systems.
- The Lunarflow interface - a React-based application that allows GUI-based interaction with Lunar workflows and their components.
The following is an overview of Lunar's directory structure denoting the frontend and backend components. The component library, Lunarverse, is organized in its own Github repository located at https://github.com/lunarbase-labs/lunarverse.
lunar
├── LICENSES
├── README.md
├── lunarcore
├── lunarflow
└── scripts
Lunarflow
Lunarflow is the web-based interface that offers effortless interaction with the universe of Lunar components.
Local installation and configuration
Lunarflow requires the following dependencies:
Once Node.js is ready the easiest way to install Lunarflow locally is using the provided installation scripts from the scripts directory (see above):
lunarflow_unix_install.sh: source lunarflow_unix_install.sh
The command above will install and make Lunarflow accessible at http://localhost:8080. To change the address or port number configuration, navigate to lunar/lunarflow/ and open .env (or create a new .env file by copying [EXAMPLE].env). Modify the corresponding lines and save the file. Upon startup Lunarflow will load the environment from .env assuming such file exists. The default values for the required variables in .env include:
| Variable name | Default value | Description |
|---|---|---|
| NEXT_PUBLIC_LUNARVERSE_ADDRESS | http://0.0.0.0:8088 | The URL of the LunarCore API. |
| NEXT_PUBLIC_HOST | http://localhost:8080 | The host URL for the frontend application (LunarVerse). |
| NEXTAUTH_URL | http://localhost:8080 | The callback URL for NextAuth authentication, normally pointing to the frontend application's host. |
The .env file should also set GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET. These will be optional soon, but in the meantime, you can generate them by configuring a project here.
Lunarcore
Lunarcore represents the backend logic of the Lunar AI expert platform. It includes core functionalities and orchestration engine (based on Prefect).
Local installation and configuration
Lunarverse is written in Python so all dependencies are managed automatically. It requires
Once Python is installed the easiest way to install Lunarcore locally is using the provided installation scripts from the scripts directory (see above):
lunarcore_unix_install.sh: source lunarcore_unix_install.sh
The commands above will make Lunarcore's REST API available on all local interfaces and port 8088. To change these settings, modify the corresponding lines in [EXAMPLE].env and rename the file to .env. Upon startup Lunarcore will load the environment from .env assuming such file exists. The default values for the required variables are in [EXAMPLE].env. These include:
| Variable name | Default value | Description |
|---|---|---|
| FLOW_STORAGE | "LOCAL" | The type of storage used by Lunarverse for various data storage needs, e.g., file uploads, workflow persistence, etc. At the moment local and S3 storage are supported |
| FLOW_STORAGE_BASE_PATH | "/tmp/lunarverse" | The root location of where data will be stored |
| LUNARVERSE_ADDRESS | "0.0.0.0" | The default API IP address (i.e., defaults to listening on all interfaces) |
| LUNARVERSE_PORT | 8088 | The default API port |
| FLOW_ROOT_DIR | "workflows" | The main directory to save workflows. The full path will be <FLOW_STORAGE_BASE_PATH>/<user_name>/<FLOW_ROOT_DIR>/<workflow_id> |
Lunar with Docker compose
Currently the docker compose setup only work with Linux or WSL2 on Windows
Windows users can install WSL2 by following the intructions here
Clone lunarecore and move inside the docker directory
git clone https://github.com/lunarbase-labs/lunarcore
cd lunarcore
cd docker
You can install git from here if it's not already available on your system.
Edit the file lunarflow_env.env then add your GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET.
GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET will be optional soon, but in the meantime, you can generate them by configuring a project here
Run the prepare.py script
python prepare.py
You can install python from here if it's not already available on your system.
Build and run the lunarbase docker compose
docker compose build && docker compose up
You can install docker with the compose plugin from here if it's not already available on your system.
Then click on this link