config: switch to Ollama-first strategy (local AI priority)

Changed API fallback chain:
- Tier 1: Ollama (local, free, always available)
- Tier 2: Gemini (free tier with quota limits)
- Tier 3: OpenRouter (cheap, flexible fallback)
- Tier 4: OpenCode (final backup)

This saves costs by using local Ollama for most tasks.
OpenRouter only used when Ollama unavailable.

backend/src/utils/gemini-fallback.js

@@ -1,10 +1,12 @@
 /**
- * Gemini API with Multi-Tier Fallback
+ * AI API Fallback System
- * Tries: Gemini → OpenRouter → OpenCode
+ * Tries: Ollama (local) → Gemini → OpenRouter → OpenCode
  */
 
 const fetch = require('node-fetch');
 
+const OLLAMA_URL = process.env.OLLAMA_URL || 'http://localhost:11434';
+const OLLAMA_MODEL = process.env.OLLAMA_MODEL || 'deepseek-v3.2:cloud';
 const GEMINI_API_KEY = process.env.GOOGLE_API_KEY;
 const OPENROUTER_API_KEY = process.env.OPENROUTER_API_KEY;
 const OPENROUTER_BASE_URL = process.env.OPENROUTER_BASE_URL || 'https://openrouter.ai/api/v1';
@@ -14,10 +16,36 @@ const OPENCODE_BASE_URL = process.env.OPENCODE_BASE_URL || 'https://api.opencode
 async function generateWithFallback(prompt, options = {}) {
   console.log('🤖 Generating content...');
   
-  // Tier 1: Try Gemini
+  // Tier 1: Try Ollama (local, free)
+  try {
+    console.log(`📍 Tier 1: Attempting Ollama (${OLLAMA_MODEL})...`);
+    const response = await fetch(`${OLLAMA_URL}/api/generate`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      timeout: 30000,
+      body: JSON.stringify({
+        model: OLLAMA_MODEL,
+        prompt: prompt,
+        stream: false,
+        temperature: options.temperature || 0.7
+      })
+    });
+    
+    if (response.ok) {
+      const data = await response.json();
+      console.log('✅ Ollama success');
+      return { success: true, provider: 'ollama', data };
+    }
+    
+    console.warn(`⚠️ Ollama error: ${response.status}, trying next...`);
+  } catch (err) {
+    console.warn(`Ollama failed: ${err.message}`);
+  }
+  
+  // Tier 2: Try Gemini
   if (GEMINI_API_KEY) {
     try {
-      console.log('📍 Tier 1: Attempting Gemini API...');
+      console.log('📍 Tier 2: Attempting Gemini API...');
       const response = await fetch(
         `https://generativelanguage.googleapis.com/v1/models/gemini-pro:generateContent?key=${GEMINI_API_KEY}`,
         {
@@ -37,7 +65,7 @@ async function generateWithFallback(prompt, options = {}) {
       }
       
       if (response.status === 429 || response.status === 403) {
-        console.warn('⚠️ Gemini quota exceeded, trying next fallback...');
+        console.warn('⚠️ Gemini quota exceeded, trying next...');
       } else {
         throw new Error(`Gemini error: ${response.status}`);
       }
@@ -46,10 +74,10 @@ async function generateWithFallback(prompt, options = {}) {
     }
   }
   
-  // Tier 2: Fallback to OpenRouter (billigare, mer flexibel)
+  // Tier 3: Fallback to OpenRouter
   if (OPENROUTER_API_KEY) {
     try {
-      console.log('📍 Tier 2: Attempting OpenRouter API...');
+      console.log('📍 Tier 3: Attempting OpenRouter API...');
       const response = await fetch(`${OPENROUTER_BASE_URL}/chat/completions`, {
         method: 'POST',
         headers: {
@@ -71,16 +99,16 @@ async function generateWithFallback(prompt, options = {}) {
         return { success: true, provider: 'openrouter', data };
       }
       
-      console.warn(`OpenRouter error: ${response.status}, trying next fallback...`);
+      console.warn(`OpenRouter error: ${response.status}, trying next...`);
     } catch (err) {
       console.warn(`OpenRouter failed: ${err.message}`);
     }
   }
   
-  // Tier 3: Fallback to OpenCode (sista försöket)
+  // Tier 4: Final fallback to OpenCode
   if (OPENCODE_API_KEY) {
     try {
-      console.log('📍 Tier 3: Attempting OpenCode API...');
+      console.log('📍 Tier 4: Attempting OpenCode API...');
       const response = await fetch(`${OPENCODE_BASE_URL}/chat/completions`, {
         method: 'POST',
         headers: {
@@ -107,12 +135,13 @@ async function generateWithFallback(prompt, options = {}) {
     }
   }
   
-  throw new Error('All generation APIs failed (Gemini → OpenRouter → OpenCode)');
+  throw new Error('All generation APIs failed (Ollama → Gemini → OpenRouter → OpenCode)');
 }
 
 module.exports = {
   generateWithFallback,
   getAvailableProviders: () => ({
+    ollama: true, // Always available locally
     gemini: !!GEMINI_API_KEY,
     openrouter: !!OPENROUTER_API_KEY,
     opencode: !!OPENCODE_API_KEY