diff --git a/docker-compose.yml b/docker-compose.yml
index df930cd..9e583c3 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -10,6 +10,7 @@ services:
       - ADMIN_PASSWORD=${ADMIN_PASSWORD:-admin}
       - OLLAMA_URL=${OLLAMA_URL:-https://ai.binarygnome.com}
       - OLLAMA_MODEL=${OLLAMA_MODEL:-gemma4:latest}
+      - OLLAMA_API_KEY=${OLLAMA_API_KEY:-}
     extra_hosts:
       - "host-gateway:host-gateway"
     networks:
diff --git a/server/routes/prompts.js b/server/routes/prompts.js
index c2bccc7..af586ce 100644
--- a/server/routes/prompts.js
+++ b/server/routes/prompts.js
@@ -3,6 +3,7 @@ import { auth } from '../middleware/auth.js'
 
 const OLLAMA_URL   = process.env.OLLAMA_URL   || 'http://localhost:11434'
 const OLLAMA_MODEL = process.env.OLLAMA_MODEL || 'llama3.2'
+const OLLAMA_KEY   = process.env.OLLAMA_API_KEY || ''   // optional Bearer token
 
 const FALLBACK = [
   // ── Original 20 ───────────────────────────────────────────
@@ -173,17 +174,23 @@ function nextFallback() {
   return fallbackQueue.pop()
 }
 
-// ── Shared Ollama call (returns parsed response or throws with detail) ────
+// ── Shared AI call — uses the OpenAI-compatible chat completions API ─────
 async function callOllama() {
-  const url = `${OLLAMA_URL}/api/generate`
-  console.log(`[prompts] calling Ollama  url=${url}  model=${OLLAMA_MODEL}`)
+  const url = `${OLLAMA_URL}/v1/chat/completions`
+  console.log(`[prompts] calling AI  url=${url}  model=${OLLAMA_MODEL}`)
+
+  const headers = { 'Content-Type': 'application/json' }
+  if (OLLAMA_KEY) headers['Authorization'] = `Bearer ${OLLAMA_KEY}`
 
   const r = await fetch(url, {
     method: 'POST',
-    headers: { 'Content-Type': 'application/json' },
+    headers,
     body: JSON.stringify({
       model: OLLAMA_MODEL,
-      prompt: 'Write one creative writing prompt for a young writer aged 10–14. Make it imaginative, specific, and intriguing — a little mysterious or adventurous. Write only the prompt itself: no introduction, no explanation, no quotation marks. Maximum two sentences.',
+      messages: [{
+        role: 'user',
+        content: 'Write one creative writing prompt for a young writer aged 10–14. Make it imaginative, specific, and intriguing — a little mysterious or adventurous. Write only the prompt itself: no introduction, no explanation, no quotation marks. Maximum two sentences.',
+      }],
       stream: false,
     }),
     signal: AbortSignal.timeout(45000),
@@ -195,7 +202,7 @@ async function callOllama() {
   }
 
   const data = await r.json()
-  const prompt = data.response?.trim()
+  const prompt = data.choices?.[0]?.message?.content?.trim()
   if (!prompt) throw new Error(`empty response — raw: ${JSON.stringify(data).slice(0, 200)}`)
   return prompt
 }