fix: cap gemini-3.1-pro maxOutputTokens and filter live models

- Reduce maxOutputTokens from 131072 to 65535 for gemini-3.1-pro-high
  and gemini-3.1-pro-low, fixing 400 "invalid argument" errors from
  Open WebUI when no max_tokens is specified (upstream limit is 65535)
- Filter non-viable models (gemini-3.1-flash-image-preview,
  gemini-2.5-flash-preview-image-generation, gemini-3-pro-high/low)
  from the live upstream API response in /api/providers/[id]/models

src/app/api/providers/[id]/models/route.ts

@@ -154,7 +154,15 @@ const PROVIDER_MODELS_CONFIG: Record<string, ProviderModelsConfigEntry> = {
     authHeader: "Authorization",
     authPrefix: "Bearer ",
     body: {},
-    parseResponse: (data) => data.models || [],
+    parseResponse: (data) => {
+      const excluded = new Set([
+        "gemini-2.5-flash-preview-image-generation",
+        "gemini-3.1-flash-image-preview",
+        "gemini-3-pro-low",
+        "gemini-3-pro-high",
+      ]);
+      return (data.models || []).filter((m: any) => !excluded.has(m.model || m.id));
+    },
   },
   openai: {
     url: "https://api.openai.com/v1/models",

src/shared/constants/modelSpecs.ts

@@ -32,7 +32,7 @@ export const MODEL_SPECS: Record<string, ModelSpec> = {
 
   // ── Gemini 3.1 Pro High ─────────────────────────────────────────
   "gemini-3.1-pro-high": {
-    maxOutputTokens: 131072,
+    maxOutputTokens: 65535,
     contextWindow: 1048576,
     defaultThinkingBudget: 24576,
     thinkingBudgetCap: 32768,
@@ -45,7 +45,7 @@ export const MODEL_SPECS: Record<string, ModelSpec> = {
 
   // ── Gemini 3.1 Pro Low ──────────────────────────────────────────
   "gemini-3.1-pro-low": {
-    maxOutputTokens: 131072,
+    maxOutputTokens: 65535,
     contextWindow: 1048576,
     defaultThinkingBudget: 8192,
     thinkingBudgetCap: 16000,