You provide a sample docker-compose (nice!), and say that is the easy way to get started, however your only step is to copy the sample docker-compose to the local-ai directory... And that example does not include any image to pull, or the Dockerfile from which to build - the way it is written I believe is only going to build the api again?

I'm sure it's an easy step, but I'm not sure what the proper step should be to get the docker compose working.

Install LocalAI inside kubernetes by using helm charts, local-ai service exposed to cluster with FQDN local-ai.default.svc.cluster.local:80 , so I construct a yaml to deploy LocalAI-fronted as follow :

# k8s deployment
apiVersion: apps/v1
kind: Deployment
metadata:
  name: localai-frontend
  namespace: default
spec:
  replicas: 1
  selector:
    matchLabels:
      app: localai-frontend
  template:
    metadata:
      labels:
        app: localai-frontend
    spec:
      containers:
      - name: localai-frontend
        image: dhruvgera/localai-frontend
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 3000
        env:
          - name: API_HOST
            value: http://local-ai.default.svc.cluster.local:80
      restartPolicy: Always
---

# k8s service
apiVersion: v1
kind: Service
metadata:
  name: localai-frontend
  namespace: default
spec:
  type: NodePort
  ports:
  - port: 80
    targetPort: 3000
    nodePort: 31000
  selector:
    app: localai-frontend

Since localai-frontend service exposed by service type Nodeport, so I visit a random node(let say it is 172.16.33.21) with port 31000. Seems it doesn't work :

1. web page create a websocket connection ws://172.16.33.21:3000/ws and failed
2. still listing models by using http://localhost:8080/v1/models

Does this project support running inside a kubernetes cluster ?

docker compose install localai and web ui:

version: '3.6'

services:
  api:
    image: quay.io/go-skynet/local-ai:latest
    ports:
      - 8080:8080
    env_file:
      - .env
    volumes:
      - ./models:/models:cached
    command: ["/usr/bin/local-ai" ]

  frontend:
    image: quay.io/go-skynet/localai-frontend:master
    ports:
      - 3000:3000

build result:

$ docker-compose up -d --pull always
[+] Running 2/2
 ✔ frontend Pulled                                                                                                 3.2s
 ✔ api Pulled                                                                                                      3.2s
[+] Building 0.0s (0/0)
[+] Running 2/0
 ✔ Container localai-frontend-1  Running                                                                           0.0s
 ✔ Container localai-api-1       Running                                                                           0.0s


$ curl http://localhost:8080/v1/models
{"object":"list","data":[{"id":"ggml-gpt4all-j","object":"model"}]}

but,the webui:

It has been a while since stabledifussion model was integrated in LocalAI.

I would like to know if it would be possible to implement this in some way, maybe a check to change the action to image generation.

The own chat UI could be kept and it could be used to list pictures generated.

I think this is somewhat easy to implement. I could check some things. Aside this, maybe whisper would be the more challenging to keep this a simple UI.

Did not check if you are already working in the image generation though.

Hi, the host ip is hard coded as local host.
Could you please move that to environment variable?
It will be more flexible to use the front end.

Thanks for the greate project!

The objective would be to get your project working as an overlay onto LocalAI running separately. I commented out the LocalAI in the docker-compose.yaml:

`❯ cat docker-compose.yaml
version: '3.6'

services:
frontend:
build:
context: .
dockerfile: Dockerfile
ports:
- 3000:3000`

❯ netstat -an | grep LISTEN
tcp46 0 0 *.3000 . LISTEN

The docker container is up and running as shown in the above image.

The API is running as an autonomous project separately and working independently. See below:

`❯ curl http://localhost:8080/v1/completions -H "Content-Type: application/json" -d '{
"model": "llama-2-7b-chat",
"prompt": "What is the expected population of Ghana by the year 2100",
"temperature": 0.7
}'

{"object":"text_completion","model":"llama-2-7b-chat","choices":[{"index":0,"finish_reason":"stop","text":"?\nlazarus May 3, 2022, 1:49pm #1\nThe population of Ghana is projected to continue growing in the coming decades. According to the United Nations Department of Economic and Social Affairs Population Division, Ghana’s population is expected to reach approximately 47 million by the year 2100. This represents a more than fivefold increase from the country’s estimated population of around 8.5 million in 2020.\nHowever, it is important to note that population projections are subject to uncertainty and can be influenced by various factors such as fertility rates, mortality rates, and migration patterns. Therefore, actual population growth may differ from projected values."}],"usage":{"prompt_tokens":0,"completion_tokens":0,"total_tokens":0}}%`

My question is how to get the "Select Model" and "Model Gallery" to effectively integrate with the LocalAI project when run separately and not directly integrated into your project? Is this possible?

I love the project concept of being able to change "model" and have "model galleries".