[feature request] Support Compose `models`

[mathieu-benoit]

Docker Compose now supports at the root level models: https://docs.docker.com/ai/compose/models-and-compose/.

Basic example:

services:
  chat-app:
    image: my-chat-app
    models:
      - llm

models:
  llm:
    model: ai/smollm2

It seems that the goal will be to add models support in the provisioners: https://github.com/score-spec/score-compose/blob/main/internal/provisioners/core.go#L85.

Open question: for patch-templates too? Not sure...

As a reference, today we can do that via a service's provider like illustrated below, but this will be deprecated soon:
https://github.com/score-spec/community-provisioners/blob/main/llm-model/score-compose/10-dmr-llm-model-via-service-provider.provisioners.yaml

[astromechza]

Yup, this should be ok to support. We just have to support it in the allowed output schema of provisioners and mix it in with the output compose file 👍

[mathieu-benoit]

This one is a good first issue that will demonstrate a very great integration of the Docker Compose project. Anyone interested?

High level guidance for the PR to open in this repo to implement this:

• Add a new Models field in this ProvisionOutput struct in there: https://github.com/score-spec/score-compose/blob/main/internal/provisioners/core.go#L85-L98
• And then take inspiration of these https://github.com/score-spec/score-compose/blob/main/internal/provisioners/core.go#L245-L262 to add the Models in the Compose's Project
• Reference of the Docker Compose's Project: https://github.com/compose-spec/compose-go/blob/main/types/project.go#L50

To test this then, this provisioner should work (model.provisioners.yaml):

- uri: template://community-provisioners/dmr-llm-model
  type: llm-model
  description: Generates the LLM model via the Docker Model Runner (DMR) provider.
  supported_params:
    - model
  outputs: |
    model: {{ .Init.model }}
    url: "http://172.17.0.1:12434/engines/v1/"
    api-key: "not-needed"
  expected_outputs:
    - model
    - url
    - api-key
  init: |
    model: {{ .Params.model | default "ai/smollm2:135M-Q4_0" }}
  models: |
    {{ .Id }}:
      model: {{ .Init.model }}

And this score.yaml:

apiVersion: score.dev/v1b1
metadata:
  name: open-webui
containers:
  open-webui:
    image: .
    variables:
      OPENAI_API_BASE_URL: "${resources.smollm2.url}"
      WEBUI_NAME: "Hello, DMR with Score Compose!"
    volumes:
      /app/backend/data:
        source: ${resources.data}
resources:
  data:
    type: volume
  gemma3:
    type: llm-model
    params:
      model: ai/gemma3:270M-UD-IQ2_XXS
  smollm2:
    type: llm-model
    params:
      model: ai/smollm2:135M-Q2_K
service:
  ports:
    tcp:
      port: 8080
      targetPort: 8080

Run these commands:

score-compose init \
  --provisioners model.provisioners.yaml

score-compose generate score.yaml \
  --image ghcr.io/open-webui/open-webui:main-slim \
  --publish 8080:open-webui:8080 \
  --output compose.yaml