No model providers found - but Docker Model Runner already setup with models

[jschmdt]

OS macOS 26.4.1 / Docker Agent v1.54.0

1. When I ran docker-agent for the first time, I was prompted to pull ai/qwen3:latest. First I declined because I already have some models installed locally via Docker Model Runner.

After declining, the CLI showed:

No model providers available.

This seems unexpected. I would assume Docker Model Runner acts as a local model provider service and that docker agent would connect to it automatically.

2. To further test the provider availability I downloaded ai/qwen3:latest – started docker-agent, but when I tried to list the models from Docker Model Runner (ctrl+m) the UI just showed …

No models available for selection

… which is a bit confusing, since at least ai/qwen3:latest should be selectable, even if its already active. And as you can see in the screenshot: Docker Model Runner is listed as the Provider dmr of the qwen model. Proof that DMR is already active as an provider but still not discovered by docker-agent when invoked with ctrl+m.

Expected Behavior

1. Models available through Docker Model Runner should be detected and listed automatically.
2. List Models works correctly when inside of docker-agents

Enhancement Request

For local agentic workflows, please consider supporting LM Studio and other local endpoints as well 😋🙏

Attachments

Docker Agent Output

$ docker agent
Model ai/qwen3:latest not found locally.
Do you want to pull it now? ([y]es/[n]o): n
No model providers available.
To fix this, you can:
  - Install Docker Model Runner: https://docs.docker.com/ai/model-runner/get-started/
  - Configure an API key for a cloud provider:
    - anthropic: ANTHROPIC_API_KEY
    - openai: OPENAI_API_KEY
    - google: GOOGLE_API_KEY (or GEMINI_API_KEY, GOOGLE_GENAI_USE_VERTEXAI)
    - mistral: MISTRAL_API_KEY
    - amazon-bedrock: AWS_ACCESS_KEY_ID (or AWS_PROFILE, AWS_ROLE_ARN, AWS_BEARER_TOKEN_BEDROCK)

---

Docker Model Runner: Models Are Available

[hamza-jeddad]

In the yaml, you can use DMR models by defining the provider as provider: dmr

models:
  default:
    provider: dmr
    model: ai/qwen3.6:latest