refactor(providers): one record per provider, one place for the key-and-URL decision (#988)

Every fact about a provider lives in one record in lib/providers.ts (SDK kind,
credential kind, env var prefix, default URL, picker group, four-language
notes, and a short model list for the six providers that cannot list models).
The thirteen tables keyed by provider name are gone. resolveEndpoint decides
whose key a call uses and where it goes; usesServerCredentials reads its
answer instead of mirroring getAIModel's branches; serverCredentialError holds
the server credential check once. The UI, the connection test and the routes
read the record's credential and sdk fields instead of comparing provider
names (44 such checks before, 6 after, in the admin panel's fixed-variable
branches).

Intended behavior changes: providers that can list their models have no
hand-written list; the settings dialog fetches the live list in the background
once credentials are entered for a provider without models and offers it
behind "Browse models". Azure has no placeholder default URL (old entries that
hold it still count as unset); the base URL field of Azure and of the new
"custom" OpenAI-compatible preset (CUSTOM_API_KEY, CUSTOM_BASE_URL) is
labelled required. The connection test answers 400 for an unknown provider
name; prototype names are refused everywhere. Provider lists in error messages
and the admin "add provider" dropdown follow the picker's order.

Also fixed, from before the refactor: the redirect guard drops x-access-code
on cross-origin redirects, and edgeOneEndpoint accepts the Origin header only
when it matches Host or X-Forwarded-Host.

Verification: a behavior snapshot of getAIModel's SDK factory calls and the
quota functions over 25 providers, 22 client override shapes and 9 server
environments (5843 rows), recorded on the old code and unchanged apart from
the intended changes; 752 unit tests, 65 provider e2e tests, tsc, biome,
next build. Reviewed by five models from one brief; their findings and the
resulting fixes are recorded in docs/plans/2026-10-provider-refactor.md.

app/api/chat/route.ts

@@ -16,7 +16,6 @@ import {
     edgeOneEndpoint,
     getAIModel,
     getServerProvider,
-    SINGLE_SYSTEM_PROVIDERS,
     supportsPromptCaching,
     usesServerCredentials,
     usesServerEndpoint,
@@ -47,6 +46,7 @@ import {
     resolveMaxOutputTokens,
     withOutputTokenLimitFallback,
 } from "@/lib/output-token-limit"
+import { PROVIDERS, type ProviderName } from "@/lib/providers"
 import {
     type FlattenedServerModel,
     findServerModelById,
@@ -173,14 +173,17 @@ async function handleChatRequest(req: Request): Promise<Response> {
     // A server model's provider comes from its config: for one set up in
     // the admin panel the header holds the provider name's slug. Without
     // either, the server's own AI_PROVIDER.
-    const isEdgeOne =
-        (serverModelConfig.provider || provider || getServerProvider()) ===
-        "edgeone"
-
-    // EdgeOne is this deployment's own function, whatever URL the request
-    // names: another host would get the user's EdgeOne cookies, and the
-    // quota counts it. Absolute, as the SDK needs.
-    if (isEdgeOne) baseUrl = edgeOneEndpoint(req)
+    const requestedProvider =
+        serverModelConfig.provider || provider || getServerProvider()
+    // A provider without credentials (EdgeOne) is this deployment's own
+    // function, whatever URL the request names: another host would get the
+    // user's EdgeOne cookies, and the quota counts it. Absolute, as the SDK
+    // needs.
+    const onSiteFunction =
+        !!requestedProvider &&
+        Object.hasOwn(PROVIDERS, requestedProvider) &&
+        PROVIDERS[requestedProvider as ProviderName].credential === "none"
+    if (onSiteFunction) baseUrl = edgeOneEndpoint(req)
 
     // Same rule as validate-model: with ALLOW_PRIVATE_URLS=false a request may
     // not point the server at a private or internal address
@@ -195,12 +198,13 @@ async function handleChatRequest(req: Request): Promise<Response> {
     const cookieHeader = req.headers.get("cookie")
 
     const clientOverrides = {
-        // Server model provider takes precedence over client header; EdgeOne
-        // named only in AI_PROVIDER is named here, for its own base URL
+        // Server model provider takes precedence over client header; the
+        // site's function named only in AI_PROVIDER is named here, for its
+        // own base URL
         provider:
             serverModelConfig.provider ||
             provider ||
-            (isEdgeOne ? "edgeone" : null),
+            (onSiteFunction ? requestedProvider : null),
         baseUrl,
         apiKey: req.headers.get("x-ai-api-key"),
         // A server model runs the model it was configured with, whatever the header says
@@ -216,7 +220,7 @@ async function handleChatRequest(req: Request): Promise<Response> {
         vertexApiKey: req.headers.get("x-vertex-api-key"),
         // Pass cookies for EdgeOne Pages authentication, and the access code,
         // which the EdgeOne function checks too
-        ...(isEdgeOne && {
+        ...(onSiteFunction && {
             headers: {
                 ...(cookieHeader && { cookie: cookieHeader }),
                 "x-access-code": req.headers.get("x-access-code") || "",
@@ -420,7 +424,7 @@ ${userInputText}
     // Also merge for OpenAI-compatible providers with custom base URLs (e.g. vLLM, LMStudio)
     // because open-source model chat templates (Qwen, Llama, etc.) typically reject multiple system messages
     const isCustomOpenAIEndpoint =
-        resolvedProvider === "openai" &&
+        PROVIDERS[resolvedProvider].sdk === "openai" &&
         !!(
             baseUrl ||
             process.env.OPENAI_BASE_URL ||
@@ -428,7 +432,7 @@ ${userInputText}
                 process.env[serverModelConfig.baseUrlEnv])
         )
     const isSingleSystemProvider =
-        SINGLE_SYSTEM_PROVIDERS.has(resolvedProvider) || isCustomOpenAIEndpoint
+        !!PROVIDERS[resolvedProvider].singleSystem || isCustomOpenAIEndpoint
 
     const selectionContext = formatSelectionContext(body.selectedCells)
     // The model reads the diagram in the compact notation it writes