use klipy instead of tenor

src/core/config.ts

@@ -16,6 +16,9 @@ interface BotConfig {
     maxTokens: number;
     temperature: number;
   };
+  klipy: {
+    apiKey: string;
+  };
   bot: {
     /** Chance of Joel responding without being mentioned (0-1) */
     freeWillChance: number;
@@ -55,15 +58,18 @@ export const config: BotConfig = {
     openRouterApiKey: getEnvOrThrow("OPENROUTER_API_KEY"),
     model: getEnvOrDefault(
       "AI_MODEL",
-      "meta-llama/llama-3.1-70b-instruct"
+      "x-ai/grok-4.1-fast"
     ),
     classificationModel: getEnvOrDefault(
       "AI_CLASSIFICATION_MODEL",
-      "meta-llama/llama-3.1-8b-instruct:free"
+      "google/gemma-3-12b-it:free" // Free model, good for simple classification
     ),
     maxTokens: parseInt(getEnvOrDefault("AI_MAX_TOKENS", "500")),
     temperature: parseFloat(getEnvOrDefault("AI_TEMPERATURE", "1.2")),
   },
+  klipy: {
+    apiKey: getEnvOrDefault("KLIPY_API_KEY", ""),
+  },
   bot: {
     freeWillChance: 0.02,
     memoryChance: 0.3,