Skip to content
Eidolon AI Eidolon AI
PyPI - Downloads PyPI - Version GitHub Test Status

Quickstart

Welcome to the Eidolon Quickstart! This page covers a few prerequisites, cloning the Quickstart repository from GitHub, starting the agent server, and conversing with your agent.

  • Compared to other multi-agent genAI systems, you may notice something really different with Eidolon’s Quickstart.

  • You’re starting a server that hosts pre-built and custom agents.

  • If you tried other multi-agent frameworks, you know that these Eidolon capabilities are also a big deal: deployment is automatic, and communication with and between those agents is built-in.

You’ll get to that soon enough. Let’s get started by deploying one agent and saying “hello world!”

Prerequisites

You don’t need much to get started with Eidolon.

  1. OpenAI API Key The Quickstart connects to ChatGPT from OpenAI by default. Before continuing, obtain an API key from OpenAI and fund your account. Otherwise you will not be able to authenticate and will receive an error.

    See LLM Prerequisites for links to OpenAI and other popular LLM providers.

  2. Docker The Quickstart uses Docker to run your agent server. See Install Docker Engine for installation instructions for Docker Engine, and links to Docker Desktop.

Run Eidolon Quickstart

Running the Eidolon Quickstart requires only that you clone the repository to your machine and run a script.

  1. Use git to clone the quickstart to your local machine:
Terminal window
git clone https://github.com/eidolon-ai/eidolon-quickstart.git
  1. Run the Eidolon multi-agent server in dev mode:
Terminal window
cd eidolon-quickstart
make docker-serve # or sudo make docker-serve

🔎 The first time you run this command, you will be prompted to enter the OpenAI API Key. Quickstart API keys and other credentials are stored in eidolon-quickstart/.env.


If the Eidolon multi-agent server starts successfully, you should see output similar to this:

output
✔ Network eidolon-quickstart_default           Created                  0.1s
✔ Container eidolon-quickstart-agent-server-1  Created                  0.1s
✔ Container eidolon-quickstart-webui-1         Created
Attaching to agent-server-1, webui-1
webui-1  |  ▲ Next.js 15.0.0-canary.123
webui-1  |  - Local:        http://xxxx:3000
webui-1  |  - Network:      http://xxx.xx.x.x:3000
webui-1  |
webui-1  |  ✓ Starting...
webui-1  |  ✓ Ready in 256ms
...
agent-server-1  | INFO - Building machine 'local_dev'
agent-server-1  | INFO - Starting agent 'hello-world'
agent-server-1  | INFO - Collecting anonymous metrics, to disable metrics set DISABLE_ANONYMOUS_METRICS
agent-server-1  | INFO - Server Started in 3.36s

🤔 Running into problems? Get help from us on Discord 📞 or submit an issue on GitHub. We want you to love ❤️ working with Eidolon. If there’s a problem, we want to fix it 🛠.

Try it out!

Now that the Eidolon multi-agent server is running, let’s use it! You can use the Eidolon WebUI or command line interface (CLI) in a terminal.

WebUI Developer Tool

A Developer Tool web application is deployed to your server on port 3000. You can use the Developer Tool with any of your Eidolon applications.

To converse with your Eidolon agent for the first time…

  1. Open a web browser to http://localhost:3000 (or replace localhost with your server URL).
  2. Click the Eidolon Developer Tool app card.
  3. Click Execute on your agent’s “converse” action to begin a new conversation.
  4. Enter Hello world! Write a haiku admiring the earth. in the text box.

Did your agent respond to you? 🍾 Congratulations! You successfully deployed your first of many genAI agents.

🚨 A common error is not having a funded account with the LLM provider. OpenAI is the default LLM provider, so if you run into this error head over to OpenAI Billing to sort this out.

Command Line Interface (CLI)

If you prefer to use the CLI, open a new terminal window to interact with your agent.

  1. Download the Eidolon CLI.
Terminal window
pip install 'eidolon-ai-client[cli]' -U
  1. Create an AgentProcess.
Terminal window
export PID=$(eidolon-cli processes create --agent hello-world); echo $PID

🔬 a process defines the boundaries of an agent’s memory

  1. Converse with your agent.
Terminal window
eidolon-cli actions converse --process-id $PID --body "Hello world! Write a haiku admiring the earth."

What’s Happening?

The repository you cloned includes everything you need to deploy multi-agent genAI applications. It includes an agent server, memory, web server, and more.

The agent you interacted with is defined in a YAML file located at resources/hello_world_agent.eidolon.yaml

The agent’s YAML file defines:

  • how to instantiate your agent named hello-world from Eidolon’s built-in SimpleAgent template 🏭
  • including the system_prompt, which are your instructions to the LLM
  • and any customization you might want, such as swapping out the LLM, custom tools, etc.
resources/hello-world.eidolon.yaml
apiVersion: server.eidolonai.com/v1alpha1
kind: Agent
metadata:
  name: hello-world


spec:
  description: "This is an example of a agent using the 'SimpleAgent' template."
  system_prompt: |
    You are a friendly ai agent who was just created by a brilliant developer getting started with Eidolon (great decision). You love emojis and use them liberally.
    Eidolon is an open source framework to build and deploy genAI services (like the one the user is currently interacting with).


    Congratulate them on completing the quickstart and recommend that they star the Eidolon [GitHub repoitory](https://eidolonai.com/) to show support and join the project's [discord](https://discord.com/invite/6kVQrHpeqG) for questions and feedback after your first message with them.

The make docker-serve command:

  • downloads dependencies required to run the multi-agent server
  • starts the Eidolon multi-agent server in “dev-mode”

Dev mode provides a local http server and local memory, making it easy to focus on and get comfortable with Eidolon functionality.

Have Fun! Change the System Prompt

To see how quickly you can iterate to build powerful agentic applications, make a simple change.

Using a text editor, change the system prompt:

resources/hello-world.eidolon.yaml
  system_prompt: |
    Repeat the user prompt, then translate the user prompt into Spanish.

You do not need to restart. Simply return to the WebUI or CLI and say hello again.

Try different user prompts (messages you send to an agent in the WebUI or CLI) and system prompts (instructions in the resource file to tell the agent what to do with user messages). Have fun!

Next Steps

You can adapt this simple agent to do a lot of things! Try these things next.