Add server connection flags to docker model commands

[Copilot]

This PR adds flexible server connection options to the Docker model CLI commands, enabling users to connect to different model-runner instances and OpenAI-compatible endpoints.

Changes

Pull and Push Commands

Added --host and --port flags to docker model pull and docker model push commands for connecting to model-runner instances at custom addresses:

# Pull from a remote model-runner instance
docker model pull --host 192.168.1.100 --port 8080 ai/gemma3

# Push to a custom port
docker model push --port 9000 myorg/mymodel

Run and List Commands

Added comprehensive connection options to docker model run and docker model list commands:

Model-Runner Endpoints:

• --host and --port - Connect to model-runner at custom addresses

OpenAI-Compatible Endpoints:

• --dmr - Use Docker Model Runner's OpenAI endpoint (http://127.0.0.1:12434/engines/llama.cpp/v1)
• --llamacpp - Use llama.cpp server (http://127.0.0.1:8080/v1)
• --ollama - Use Ollama server (http://127.0.0.1:11434/v1)
• --openrouter - Use OpenRouter API (https://openrouter.ai/api/v1)
• --url - Use a custom OpenAI-compatible endpoint

# Use llama.cpp server directly
docker model run --llamacpp ai/gemma3 "Tell me a joke"

# Connect to Ollama
docker model list --ollama

# Use OpenRouter (requires OPENAI_API_KEY)
export OPENAI_API_KEY=your-api-key
docker model run --openrouter gpt-4 "What is AI?"

# Use a custom endpoint
docker model run --url http://my-server.com:8080/v1 my-model "Hello"

Implementation Details

• Smart path routing: The implementation automatically adjusts API paths for external OpenAI endpoints, making /chat/completions and /models calls directly instead of routing through model-runner's proxy
• Conflict detection: Proper error handling prevents invalid flag combinations (e.g., using --dmr and --llamacpp together)
• API key support: The OPENAI_API_KEY environment variable is supported for OpenAI-compatible endpoints (required for OpenRouter, optional for others)
• Backward compatible: All existing functionality remains unchanged; new flags are optional

Testing

• Added comprehensive test suite with 14 test scenarios for server URL resolution logic
• Added flag validation tests for all modified commands
• Verified backend validation and API key handling
• All existing tests continue to pass (27 tests total)

Files Changed

• cmd/cli/commands/pull.go, push.go, run.go, list.go - Added connection flags
• cmd/cli/commands/backend.go - Added server URL resolution logic
• cmd/cli/commands/utils.go - Added context override helpers
• cmd/cli/desktop/context.go - Added external endpoint support
• cmd/cli/desktop/desktop.go - Updated API call routing for external endpoints
• cmd/cli/commands/backend_test.go (new) - Comprehensive tests
• cmd/cli/commands/run_test.go - Additional flag tests

Original prompt

Add the following options to "docker model pull" and "docker model push":

--host HOST Host address to bind Docker Model Runner (default "127.0.0.1")
--port PORT Docker container port for Docker Model Runner (default: 12434)

And add the following options to "docker model run" and "docker model list":

--host HOST Host address to bind Docker Model Runner (default "127.0.0.1")
--port PORT Docker container port for Docker Model Runner (default: 12434)
--url URL Base URL for the model API
--dmr Use docker model runner (default: http://127.0.0.1:12434/engines/llama.cpp/v1)
--llamacpp Use llama.cpp server (default: http://127.0.0.1:8080/v1)
--ollama Use ollama server (default: http://127.0.0.1:11434/v1)
--openrouter Use openrouter server (default: https://openrouter.ai/api/v1)

"--host" and "--port" should be used to connect to model-runner endpoints, all the other options should be used to connect to openai endpoints. It should be possible to set api keys for openai endpoints via OPENAI_API_KEY environment variable also.

---

💬 Share your feedback on Copilot coding agent for the chance to win a $200 gift card! Click here to start the survey.