diff --git a/app/api/inference/route.ts b/app/api/inference/route.ts index 1390bb6..d29e9f9 100644 --- a/app/api/inference/route.ts +++ b/app/api/inference/route.ts @@ -180,3 +180,29 @@ export async function POST(request: Request) { return NextResponse.json({ ok: false, error: message }, { status: /required|unsupported|must|URL/.test(message) ? 400 : 500 }) } } + +export async function DELETE(request: Request) { + try { + const body = await request.json().catch(() => ({})) + const url = new URL(request.url) + const name = validateName(body?.name ?? url.searchParams.get("name")) + const result = await runOpenShell(["provider", "delete", name]) + const [inference, providers] = await Promise.all([ + runOpenShell(["inference", "get"]), + readProviders(), + ]) + + return NextResponse.json({ + ok: true, + provider: name, + gateway: parseRoute(inference.stdout, "Gateway inference"), + system: parseRoute(inference.stdout, "System inference"), + providers, + stdout: result.stdout, + stderr: result.stderr, + }) + } catch (error) { + const message = error instanceof Error ? error.message : "Failed to delete inference provider" + return NextResponse.json({ ok: false, error: message }, { status: /required|must/.test(message) ? 400 : 500 }) + } +} diff --git a/app/components/InferenceEndpointPanel.tsx b/app/components/InferenceEndpointPanel.tsx index c4ebf06..e467024 100644 --- a/app/components/InferenceEndpointPanel.tsx +++ b/app/components/InferenceEndpointPanel.tsx @@ -68,6 +68,7 @@ export default function InferenceEndpointPanel() { const [ollamaModels, setOllamaModels] = useState([]) const [ollamaLoading, setOllamaLoading] = useState(false) const [ollamaMessage, setOllamaMessage] = useState("") + const [deletingProvider, setDeletingProvider] = useState(null) const activeProvider = useMemo( () => providers.find((provider) => provider.name === gateway.provider) ?? null, @@ -140,6 +141,17 @@ export default function InferenceEndpointPanel() { setMessage(`Loaded ${provider.name}. Enter a model and save to make it active.`) } + function useNvidiaInferencePreset() { + setName("nvidia-inference") + setType("openai") + setModel("us/aws/anthropic/bedrock-claude-opus-4-6") + setBaseUrl("https://inference-api.nvidia.com/v1") + setCredentialKey("NVIDIA_API_KEY") + setRoute("gateway") + setNoVerify(true) + setMessage("NVIDIA inference preset loaded. Enter an API key, then save the endpoint.") + } + async function saveEndpoint() { try { setSaving(true) @@ -174,6 +186,31 @@ export default function InferenceEndpointPanel() { } } + async function deleteProvider(providerName: string | null) { + if (!providerName) return + if (!window.confirm(`Delete inference provider '${providerName}'?`)) return + try { + setDeletingProvider(providerName) + setMessage("") + const response = await fetch("/api/inference", { + method: "DELETE", + headers: { "Content-Type": "application/json" }, + body: JSON.stringify({ name: providerName }), + }) + const data = await response.json() + if (!response.ok) throw new Error(data.error || "Failed to delete inference provider") + setGateway(data.gateway ?? gateway) + setSystem(data.system ?? system) + setProviders(Array.isArray(data.providers) ? data.providers : []) + if (name === providerName) setName("") + setMessage(`Deleted inference provider '${providerName}'.`) + } catch (error) { + setMessage(error instanceof Error ? error.message : "Failed to delete inference provider") + } finally { + setDeletingProvider(null) + } + } + return (
@@ -296,6 +333,9 @@ export default function InferenceEndpointPanel() { + {message &&

{message}

}
@@ -307,13 +347,18 @@ export default function InferenceEndpointPanel() { ) : (
{providers.map((provider) => ( - +
+ + +
))}
)} diff --git a/tests/sandbox-lifecycle-check.mjs b/tests/sandbox-lifecycle-check.mjs index 550fbf1..8835039 100644 --- a/tests/sandbox-lifecycle-check.mjs +++ b/tests/sandbox-lifecycle-check.mjs @@ -69,9 +69,13 @@ assert.match(telemetryRouteSource, /const names = parseOpenShellSandboxNames\(sa assert.match(telemetryRouteSource, /names\.length > 0[\s\S]*execNemoclaw\(\["list"\]\)[\s\S]*execNemoclaw\(\["status"\]\)/, 'telemetry must skip NemoClaw list/status when OpenShell reports zero live sandboxes') assert.match(inferenceRouteSource, /\["provider", "create", "--name", name, "--type", type\]/, 'inference route must create OpenShell providers') assert.match(inferenceRouteSource, /\["provider", "update", name\]/, 'inference route must update existing OpenShell providers') +assert.match(inferenceRouteSource, /\["provider", "delete", name\]/, 'inference route must delete OpenShell providers') assert.match(inferenceRouteSource, /\["inference", "set", "--provider", name, "--model", model\]/, 'inference route must set the active OpenShell inference route') assert.match(inferenceRouteSource, /OPENAI_BASE_URL=\$\{baseUrl\}/, 'inference route must pass endpoint URL as provider config') assert.match(inferencePanelSource, /INFERENCE ENDPOINTS/, 'settings UI must expose inference endpoint configuration') +assert.match(inferencePanelSource, /https:\/\/inference-api\.nvidia\.com\/v1/, 'settings UI must provide the NVIDIA inference API preset') +assert.match(inferencePanelSource, /bedrock-claude-opus-4-6/, 'settings UI must include the requested NVIDIA inference model preset') +assert.match(inferencePanelSource, /method: "DELETE"/, 'settings UI must expose provider deletion') assert.match(inferencePanelSource, /\/api\/ollama\/models/, 'settings UI must poll local Ollama models') assert.match(inferencePanelSource, /window\.setInterval\(\(\) => loadOllamaModels\(\{ silent: true \}\), 10000\)/, 'settings UI must autopoll Ollama while selected') assert.match(ollamaModelsRouteSource, /\/api\/tags/, 'Ollama models route must query the Ollama tags API')