Restore create-time inference selection

This commit is contained in:
mmckeen-nv
2026-04-30 10:13:23 -07:00
parent fda3cbb056
commit 443eba4ae1
3 changed files with 147 additions and 5 deletions
+60 -1
View File
@@ -57,6 +57,60 @@ type NemoClawCreateCommand = {
mode: "cli-onboard" | "legacy-setup"
}
type CreateInferenceMode = "auto" | "vllm" | "nim"
type CreateInferenceSettings = {
mode: CreateInferenceMode
model: string
hasApiKey: boolean
envSummary: string[]
}
function parseCreateInferenceSettings(body: any): CreateInferenceSettings {
const raw = body?.createInference && typeof body.createInference === "object" ? body.createInference : {}
const requestedMode = typeof raw.mode === "string" ? raw.mode : "auto"
const mode: CreateInferenceMode = requestedMode === "vllm" || requestedMode === "nim" ? requestedMode : "auto"
const model = typeof raw.model === "string" ? raw.model.trim() : ""
const apiKey = typeof raw.apiKey === "string" ? raw.apiKey.trim() : ""
return {
mode,
model,
hasApiKey: Boolean(apiKey),
envSummary: [],
}
}
function applyCreateInferenceEnv(env: NodeJS.ProcessEnv, settings: CreateInferenceSettings, body: any) {
if (settings.mode === "auto") return settings
env.NEMOCLAW_EXPERIMENTAL = "1"
settings.envSummary.push("NEMOCLAW_EXPERIMENTAL=1")
if (settings.mode === "vllm") {
env.NEMOCLAW_PROVIDER = "vllm"
settings.envSummary.push("NEMOCLAW_PROVIDER=vllm")
}
if (settings.mode === "nim") {
env.NEMOCLAW_PROVIDER = "nim-local"
settings.envSummary.push("NEMOCLAW_PROVIDER=nim-local")
const apiKey = typeof body?.createInference?.apiKey === "string" ? body.createInference.apiKey.trim() : ""
if (apiKey) {
env.NVIDIA_API_KEY = apiKey
env.NEMOCLAW_PROVIDER_KEY = apiKey
settings.envSummary.push("NVIDIA_API_KEY=<provided>", "NEMOCLAW_PROVIDER_KEY=<provided>")
}
}
if (settings.model) {
env.NEMOCLAW_MODEL = settings.model
settings.envSummary.push(`NEMOCLAW_MODEL=${settings.model}`)
}
return settings
}
function buildNemoClawCreateCommand(): NemoClawCreateCommand {
if (NEMOCLAW_BIN && commandExists(NEMOCLAW_BIN)) {
const args = ["onboard", "--non-interactive", "--recreate-sandbox", "--yes-i-accept-third-party-software"]
@@ -331,7 +385,8 @@ export async function POST(request: Request) {
const sandboxName = validateSandboxName(typeof body?.sandboxName === "string" ? body.sandboxName.trim() : "")
const policy = body?.policy && typeof body.policy === "object" ? body.policy : null
const enableTailscale = Boolean(body?.enableTailscale)
console.log(`[sandbox/create] request:parsed sandbox=${sandboxName} blueprint=${blueprint} enableTailscale=${enableTailscale} elapsedMs=${elapsedMs(requestStartedAt)}`)
const createInference = parseCreateInferenceSettings(body)
console.log(`[sandbox/create] request:parsed sandbox=${sandboxName} blueprint=${blueprint} enableTailscale=${enableTailscale} inferenceMode=${createInference.mode} elapsedMs=${elapsedMs(requestStartedAt)}`)
if (!blueprint) {
return NextResponse.json({ ok: false, error: "blueprint is required" }, { status: 400 })
@@ -353,6 +408,8 @@ export async function POST(request: Request) {
env.NVIDIA_API_KEY = env.NVIDIA_API_KEY || "optional-local-mode"
}
applyCreateInferenceEnv(env, createInference, body)
const result = await runCommand(
createCommand.file,
createCommand.mode === "legacy-setup" ? [...createCommand.args, sandboxName] : createCommand.args,
@@ -369,6 +426,7 @@ export async function POST(request: Request) {
blueprint,
sandboxName,
enableTailscale,
createInference,
}, { status: 500 })
}
@@ -393,6 +451,7 @@ export async function POST(request: Request) {
verification,
mode: "nemoclaw-blueprint",
enableTailscale,
createInference,
createCommand,
hostPath: HOST_PATH,
readiness: {