Normalize inference endpoint URLs

This commit is contained in:
mmckeen-nv
2026-04-29 13:01:59 -07:00
parent 10f5a0b6aa
commit e55dc7b9e0
5 changed files with 63 additions and 7 deletions
+7 -2
View File
@@ -1,6 +1,7 @@
import { NextResponse } from "next/server"
import { execFile } from "node:child_process"
import { promisify } from "node:util"
import { normalizeInferenceBaseUrlForGateway } from "@/app/lib/inferenceEndpointUrl"
import { OPENSHELL_BIN, hostCommandEnv } from "@/app/lib/hostCommands"
const execFileAsync = promisify(execFile)
@@ -144,6 +145,10 @@ export async function POST(request: Request) {
const model = optionalString(body?.model)
const baseUrl = optionalString(body?.baseUrl)
const type = normalizeProviderType(requestedType, name, baseUrl)
const normalizedBaseUrl = normalizeInferenceBaseUrlForGateway(
baseUrl,
type === "anthropic" ? "anthropic" : "openai",
)
const apiKey = optionalString(body?.apiKey)
const credentialKey = optionalString(body?.credentialKey) || "OPENAI_API_KEY"
const makeActive = body?.makeActive !== false
@@ -153,7 +158,7 @@ export async function POST(request: Request) {
if (makeActive && !model) throw new Error("model is required when making the endpoint active")
if (!/^[A-Z_][A-Z0-9_]*$/.test(credentialKey)) throw new Error("credential key must look like an environment variable name")
if (baseUrl) new URL(baseUrl)
if (normalizedBaseUrl) new URL(normalizedBaseUrl)
const existingNames = (await readProviders()).map((provider) => provider.name).filter(Boolean)
const providerArgs = existingNames.includes(name)
@@ -163,7 +168,7 @@ export async function POST(request: Request) {
const providerEnv = apiKey ? { [credentialKey]: apiKey } : {}
if (!existingNames.includes(name) && !apiKey) providerArgs.push("--from-existing")
if (apiKey) providerArgs.push("--credential", credentialKey)
if (baseUrl) providerArgs.push("--config", `OPENAI_BASE_URL=${baseUrl}`)
if (normalizedBaseUrl) providerArgs.push("--config", `OPENAI_BASE_URL=${normalizedBaseUrl}`)
const providerResult = await runOpenShell(providerArgs, providerEnv)
let inferenceResult = null
+3 -1
View File
@@ -4,6 +4,7 @@ import { existsSync } from "node:fs"
import { readFile, writeFile } from "node:fs/promises"
import path from "node:path"
import { promisify } from "node:util"
import { normalizeInferenceBaseUrlForGateway } from "@/app/lib/inferenceEndpointUrl"
import { inspectSandbox } from "@/app/lib/openshellHost"
import {
commandExists,
@@ -127,7 +128,8 @@ function applyConfiguredInferenceToOnboardEnv(env: NodeJS.ProcessEnv, route: Con
}
if (provider === "custom") {
const endpointUrl = process.env.NEMOCLAW_ENDPOINT_URL || process.env.OPENAI_BASE_URL || process.env.VLLM_BASE_URL || ""
if (endpointUrl) env.NEMOCLAW_ENDPOINT_URL = endpointUrl
const normalizedEndpointUrl = normalizeInferenceBaseUrlForGateway(endpointUrl, "openai")
if (normalizedEndpointUrl) env.NEMOCLAW_ENDPOINT_URL = normalizedEndpointUrl
const providerKey = process.env.NEMOCLAW_PROVIDER_KEY || process.env.OPENAI_API_KEY || process.env.VLLM_API_KEY || ""
if (providerKey) env.NEMOCLAW_PROVIDER_KEY = providerKey
}