Skip to main content

Getting Started

Test LMOS locally​

The LMOS Demo serves as a starting point for testing LMOS. While we are still in the process of migrating projects to Open Source and adopting Open Standards, the core concepts are already available for testing.
The LMOS Demo launches a container that internally sets up Kubernetes (Minikube), along with Istio, Kiali, Grafana, and Prometheus, into which the LMOS components are installed.

Prerequisites:

Before you begin, ensure the following tools are installed and running on your local machine:

1. Open the Repository in a Dev Container​

  1. Clone the repository:

    git clone https://github.com/lmos-ai/lmos-demo.git
    cd lmos-demo

  2. Open the repository in Visual Studio Code:

  3. Open the Command Palette (F1 or Ctrl+Shift+P on Windows, Shift+Command+P on Mac) and select Remote-Containers: Reopen in Container. This will build and open the repository in a Docker-based development container. (Please note: a. If you are unable to find the option 'Remote-Containers: Reopen in Container' in Command Palette then use 'Dev Containers: Rebuild Container' b. Default memory allocated is 8 gigs, if you are facing any container boot issue related to memory then you may reduce this allocated memory by updating memory parameter available in 'runArgs' in 'devcontainer.json' file)

2. Set OpenAI Connection Details​

Once inside the development container, set up the necessary environment variables for OpenAI API access in the .env file. This OpenAPI access is used by the lmos-runtime and the agents.

OPENAI_APIKEY="<your-openai-api-key>"
OPENAI_MODELNAME="gpt-4o-mini"
OPENAI_URL="https://api.openai.com"

3. Check the Setup​

To verify the installation of LMOS, run:

kubectl get pods

Output:

NAME                               READY   STATUS    RESTARTS   AGE
lmos-operator-c45887647-bcwf8      2/2     Running   0          4m16s
lmos-runtime-85654bc6bc-chvrj      2/2     Running   0          4m15s

The status has to be 2/2 Running.

Two agents have been installed, you can list them with

kubectl get agents

Output:

NAME                AGE
arc-news-agent      2m34s
arc-weather-agent   2m35s

One channel has been defined, using the capability of the weather-agent.

You can list available channels with the following command:

kubectl get channels

Output:

NAME               RESOLVE_STATUS
acme-web-stable    RESOLVED

The RESOLVE_STATUS of the channel has to be RESOLVED, which means the required capabilities have been resolved. If the status is UNRESOLVED, you can check the reason with:

kubectl get channel acme-web-stable -o yaml

Output:

apiVersion: lmos.ai/v1
kind: Channel
metadata:
  name: acme-web-stable
  labels:
    tenant: acme
    channel: web
    version: 1.0.0
    subset: stable
spec:
  requiredCapabilities:
    - name: get-weather-forecast
      version: ">=1.0.4"

You can list the resolved channelroutings with:

kubectl get channelroutings

And look at a specific channel routing with:

kubectl get channelrouting acme-web-stable -o yaml

4. Access Kiali and Grafana​

To visualize your setup, various ports have been forwarded for LMOS, Kiali, Prometheus, Jaeger, Grafana and ArgoCD. You can access these tools at

5. Execute a POST request​

You can use Postman or the test_runtime.sh script to send a test request to the LMOS runtime. The lmos-runtime is uses the lmos-router to route the request to the appropriate agent.

To test the weather agent, run:

./test_runtime.sh

Output:

{"content":"The weather in London is 21 degrees."}

You will see that the weather-agent has responded.

Deploy LMOS on your Kubernetes cluster​

This guides provides instructions to install lmos-operator and lmos-runtime on your Kubernetes cluster.

Prerequisites:

Before proceeding with the installation, ensure you have the following prerequisites:

  • Kubernetes cluster (v1.19 or newer).
  • Helm installed (v3 or newer).
  • Access to the OpenAI API.
  • The OPENAI_API_KEY and OPENAI_API_URL values should be available.

1. Install lmos-operator​

To install lmos-operator using Helm, run the following command:

helm upgrade --install lmos-operator oci://ghcr.io/lmos-ai/lmos-operator-chart \
  --version 0.0.4-SNAPSHOT

2. Create Kubernetes Secret for OpenAI​

Next, you need to create a Kubernetes secret that contains your OpenAI API key. Replace "$OPENAI_API_KEY" with your actual OpenAI API key.

kubectl create secret generic lmos-runtime --from-literal=OPENAI_API_KEY="$OPENAI_API_KEY"

3. Install lmos-runtime​

Now, install lmos-runtime using Helm. Replace the environment variables with the appropriate values:

  • "$OPENAI_API_URL": Your OpenAI API URL (e.g., https://api.openai.com).
  • GPT4o-mini: The desired OpenAI model (in this case, GPT4o-mini).
helm upgrade --install lmos-runtime oci://ghcr.io/lmos-ai/lmos-runtime-chart \
  --version 0.0.8-SNAPSHOT \
  --set openaiApiUrl="$OPENAI_API_URL" \
  --set openaiApiModel=GPT4o-mini \
  --set agentRegistryUrl=http://lmos-operator.default.svc.cluster.local:8080

4. Verifying Installation​

To ensure both components are installed and running correctly, use the following commands to check the status of the pods:

kubectl get pods

You should see both lmos-operator and lmos-runtime pods in a running state.

Develop your own agent​

With ARC, we offer a Kotlin-based framework for developing agents. ARC comes with its own dedicated documentation.