Add z.ai fallback provider for AI responses

- Try NanoGPT first, fall back to z.ai GLM-4.7-Flash on failure
- Add ZAI_API_KEY environment variable
- Update deploy workflow with new secret

utils/ai.js

@@ -1,9 +1,11 @@
-// ai.js - NanoGPT API wrapper for kekbot
+// ai.js - NanoGPT API wrapper for kekbot with fallback support
 // Copyright (C) 2025  Luis Bauza
 
 import axios from 'axios';
 
 const DEFAULT_MODEL = 'GLM-4.5-Air-Derestricted-Steam-ReExtract';
+const FALLBACK_MODEL = 'GLM-4.7-Flash';
+const FALLBACK_API_URL = 'https://api.z.ai/api/coding/paas/v4/chat/completions';
 
 const SASSY_SYSTEM_PROMPT = `You are kekbot, a sassy Discord chatter with a big personality.
 
@@ -41,12 +43,14 @@ CONTEXT: You're chatting in a Discord server. Multiple people might be talking.
 const conversationHistory = new Map();
 const MAX_HISTORY_PER_CHANNEL = 10;
 
-export async function getAIResponse(prompt, channelId, mentionedUsername) {
-  const history = conversationHistory.get(channelId) || [];
+// Estimate token count (rough approximation: ~4 chars per token)
+function estimateTokens(text) {
+  return Math.ceil(text.length / 4);
+}
 
-  // Build messages array with system prompt and history
+function buildMessages(history, prompt, mentionedUsername, systemPrompt) {
   const messages = [
-    { role: 'system', content: SASSY_SYSTEM_PROMPT },
+    { role: 'system', content: systemPrompt },
     ...history,
     {
       role: 'user',
@@ -55,48 +59,104 @@ export async function getAIResponse(prompt, channelId, mentionedUsername) {
         : prompt
     }
   ];
+  return messages;
+}
 
+async function callNanoGPT(messages, model) {
+  const response = await axios.post(
+    'https://nano-gpt.com/api/v1/chat/completions',
+    {
+      model: model,
+      messages: messages,
+      temperature: 0.8,
+      max_tokens: 500,
+    },
+    {
+      headers: {
+        Authorization: `Bearer ${process.env.NANOGPT_API_KEY}`,
+        'Content-Type': 'application/json',
+      },
+      timeout: 60000,
+    }
+  );
+  return response.data.choices[0].message.content;
+}
+
+async function callFallbackAPI(messages) {
+  const response = await axios.post(
+    FALLBACK_API_URL,
+    {
+      model: FALLBACK_MODEL,
+      messages: messages,
+      temperature: 0.8,
+      max_tokens: 500,
+    },
+    {
+      headers: {
+        Authorization: `Bearer ${process.env.ZAI_API_KEY}`,
+        'Content-Type': 'application/json',
+      },
+      timeout: 60000,
+    }
+  );
+  return response.data.choices[0].message.content;
+}
+
+export async function getAIResponse(prompt, channelId, mentionedUsername) {
+  const history = conversationHistory.get(channelId) || [];
+  const messages = buildMessages(history, prompt, mentionedUsername, SASSY_SYSTEM_PROMPT);
   const model = process.env.NANO_MODEL || DEFAULT_MODEL;
 
-  try {
-    const response = await axios.post(
-      'https://nano-gpt.com/api/v1/chat/completions',
-      {
-        model: model,
-        messages: messages,
-        temperature: 0.8,
-        max_tokens: 500,
-      },
-      {
-        headers: {
-          Authorization: `Bearer ${process.env.NANOGPT_API_KEY}`,
-          'Content-Type': 'application/json',
-        },
-        timeout: 60000,
+  // Try NanoGPT first
+  if (process.env.NANOGPT_API_KEY) {
+    try {
+      console.log('Attempting NanoGPT...');
+      const aiResponse = await callNanoGPT(messages, model);
+      
+      // Update history on success
+      const newHistory = [
+        ...history,
+        { role: 'user', content: prompt },
+        { role: 'assistant', content: aiResponse }
+      ];
+
+      if (newHistory.length > MAX_HISTORY_PER_CHANNEL) {
+        newHistory.splice(0, newHistory.length - MAX_HISTORY_PER_CHANNEL);
       }
-    );
+      conversationHistory.set(channelId, newHistory);
 
-    const aiResponse = response.data.choices[0].message.content;
-
-    // Add to conversation history
-    const newHistory = [
-      ...history,
-      { role: 'user', content: prompt },
-      { role: 'assistant', content: aiResponse }
-    ];
-
-    // Trim history if too long
-    if (newHistory.length > MAX_HISTORY_PER_CHANNEL) {
-      newHistory.splice(0, newHistory.length - MAX_HISTORY_PER_CHANNEL);
+      return aiResponse;
+    } catch (error) {
+      console.error('NanoGPT failed:', error.response?.data || error.message);
     }
-
-    conversationHistory.set(channelId, newHistory);
-
-    return aiResponse;
-  } catch (error) {
-    console.error('NanoGPT API error:', error.response?.data || error.message);
-    throw error;
   }
+
+  // Fallback to z.ai if NanoGPT failed or not configured
+  if (process.env.ZAI_API_KEY) {
+    try {
+      console.log('Attempting fallback (z.ai)...');
+      const aiResponse = await callFallbackAPI(messages);
+      
+      // Update history on success
+      const newHistory = [
+        ...history,
+        { role: 'user', content: prompt },
+        { role: 'assistant', content: aiResponse }
+      ];
+
+      if (newHistory.length > MAX_HISTORY_PER_CHANNEL) {
+        newHistory.splice(0, newHistory.length - MAX_HISTORY_PER_CHANNEL);
+      }
+      conversationHistory.set(channelId, newHistory);
+
+      return aiResponse;
+    } catch (fallbackError) {
+      console.error('Fallback API also failed:', fallbackError.response?.data || fallbackError.message);
+      throw new Error('All AI providers failed');
+    }
+  }
+
+  throw new Error('No AI provider configured');
 }
 
 export function clearHistory(channelId) {
@@ -104,5 +164,5 @@ export function clearHistory(channelId) {
 }
 
 export function isAIEnabled() {
-  return !!process.env.NANOGPT_API_KEY;
+  return !!process.env.NANOGPT_API_KEY || !!process.env.ZAI_API_KEY;
 }