feat(providers): add Docker Model Runner provider

Add model-runner as a built-in inference provider so agents running
inside OpenShell sandboxes can route inference requests to the local
Docker Model Runner daemon via model-runner.docker.internal:80.

No credentials are required; Docker Model Runner is accessed over the
Docker-internal network without authentication by default.

Author: ericcurtin

crates/openshell-providers/src/lib.rs

@@ -91,6 +91,7 @@ impl ProviderRegistry {
         registry.register(providers::nvidia::NvidiaProvider);
         registry.register(providers::gitlab::GitlabProvider);
         registry.register(providers::github::GithubProvider);
+        registry.register(providers::model_runner::ModelRunnerProvider);
         registry.register(providers::outlook::OutlookProvider);
         registry
     }
@@ -153,6 +154,7 @@ pub fn normalize_provider_type(input: &str) -> Option<&'static str> {
         "nvidia" => Some("nvidia"),
         "gitlab" | "glab" => Some("gitlab"),
         "github" | "gh" => Some("github"),
+        "model-runner" | "model_runner" => Some("model-runner"),
         "outlook" => Some("outlook"),
         _ => None,
     }

crates/openshell-providers/src/profiles.rs

@@ -21,6 +21,7 @@ const BUILT_IN_PROFILE_YAMLS: &[&str] = &[
     include_str!("../../../providers/copilot.yaml"),
     include_str!("../../../providers/github.yaml"),
     include_str!("../../../providers/gitlab.yaml"),
+    include_str!("../../../providers/model-runner.yaml"),
     include_str!("../../../providers/nvidia.yaml"),
     include_str!("../../../providers/openai.yaml"),
     include_str!("../../../providers/opencode.yaml"),

crates/openshell-providers/src/providers/mod.rs

@@ -8,6 +8,7 @@ pub mod copilot;
 pub mod generic;
 pub mod github;
 pub mod gitlab;
+pub mod model_runner;
 pub mod nvidia;
 pub mod openai;
 pub mod opencode;

crates/openshell-providers/src/providers/model_runner.rs

@@ -0,0 +1,35 @@
+// SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+use crate::{DiscoveredProvider, ProviderError, ProviderPlugin};
+
+pub struct ModelRunnerProvider;
+
+impl ProviderPlugin for ModelRunnerProvider {
+    fn id(&self) -> &'static str {
+        "model-runner"
+    }
+
+    fn discover_existing(&self) -> Result<Option<DiscoveredProvider>, ProviderError> {
+        Ok(Some(DiscoveredProvider::default()))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::ModelRunnerProvider;
+    use crate::ProviderPlugin;
+
+    #[test]
+    fn model_runner_provider_id_is_correct() {
+        assert_eq!(ModelRunnerProvider.id(), "model-runner");
+    }
+
+    #[test]
+    fn model_runner_discover_returns_default_provider() {
+        let result = ModelRunnerProvider
+            .discover_existing()
+            .expect("discovery should succeed");
+        assert!(result.is_some());
+    }
+}

providers/model-runner.yaml

@@ -0,0 +1,15 @@
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+id: model-runner
+display_name: Docker Model Runner
+description: Local AI inference via Docker Model Runner
+category: inference
+inference_capable: true
+endpoints:
+  - host: model-runner.docker.internal
+    port: 80
+    protocol: rest
+    access: read-write
+    enforcement: enforce
+binaries: [/usr/local/bin/docker]