diff --git a/src/app/manufacturers/page.tsx b/src/app/manufacturers/page.tsx
index d2798c2..7627349 100644
--- a/src/app/manufacturers/page.tsx
+++ b/src/app/manufacturers/page.tsx
@@ -82,9 +82,9 @@ export default async function ManufacturersIndex() {
           </p>
           <p className="mt-2">
             Are you a manufacturer and want to update your listing?{" "}
-            <Link href="/contact" className="text-[#3b82f6] hover:underline">
+            <a href="mailto:editors@broadcastbeat.com?subject=Manufacturer%20Directory%20Update" className="text-[#3b82f6] hover:underline">
               Contact our editors
-            </Link>
+            </a>
             .
           </p>
         </section>
diff --git a/src/lib/ai/hybridRouter.ts b/src/lib/ai/hybridRouter.ts
index 22ba3b1..8603145 100644
--- a/src/lib/ai/hybridRouter.ts
+++ b/src/lib/ai/hybridRouter.ts
@@ -197,12 +197,17 @@ export async function hybridAI(
   const endpoint = getOllamaEndpoint();
   const ollamaModel = getOllamaModel(taskType);
 
-  // Skip Ollama for low-priority tasks when no endpoint configured
-  // or when priority is 5-6 (translation/classification) — route direct to Anthropic
-  // to avoid queuing behind interactive sessions
-  const skipOllama = !endpoint || !ollamaModel || priority >= 5;
+  if (!endpoint || !ollamaModel) {
+    throw new Error('Ollama not configured (OLLAMA_ENDPOINT or model missing) — refusing to fall back to paid API');
+  }
 
-  if (!skipOllama) {
+  // Retry-with-backoff against Ollama. Per the no-paid-AI policy we never
+  // fall back to Anthropic, so the caller gets a clear error and is expected
+  // to surface its own graceful fallback (e.g. non-AI default).
+  const RETRY_DELAYS_MS = [0, 3000, 12000];
+  let lastErr: any = null;
+  for (let i = 0; i < RETRY_DELAYS_MS.length; i++) {
+    if (RETRY_DELAYS_MS[i] > 0) await new Promise((r) => setTimeout(r, RETRY_DELAYS_MS[i]));
     const start = Date.now();
     try {
       const text = await callOllama(messages, ollamaModel, endpoint, options);
@@ -210,24 +215,18 @@ export async function hybridAI(
       logHealthEvent({ provider: 'ollama', success: true, latencyMs, model: ollamaModel, priority });
       return { text, provider: 'ollama', model: ollamaModel, latencyMs };
     } catch (err: any) {
+      lastErr = err;
       const latencyMs = Date.now() - start;
       logHealthEvent({
         provider: 'ollama', success: false, latencyMs, model: ollamaModel, priority,
-        errorMessage: err?.message, isFailover: true,
+        errorMessage: err?.message,
       });
-      // Fall through to Anthropic
+      // Retry-eligible if message looks like 503/busy/timeout; otherwise bail.
+      if (!/503|busy|pending|queue|timeout|empty|abort/i.test(String(err?.message || ''))) break;
     }
   }
 
-  // Anthropic fallback
-  const start = Date.now();
-  const { text, model } = await callAnthropic(messages, options);
-  const latencyMs = Date.now() - start;
-  logHealthEvent({
-    provider: 'anthropic', success: true, latencyMs, model, priority,
-    isFailover: !skipOllama,
-  });
-  return { text, provider: 'anthropic', model, latencyMs };
+  throw lastErr || new Error('Ollama failed after retries');
 }
 
 // ─── Streaming variant (for AI Console) ──────────────────────────────────────
@@ -272,55 +271,9 @@ export async function hybridAIStream(
     }
   }
 
-  // Anthropic streaming fallback
-  const apiKey = getAnthropicKey();
-  if (!apiKey) throw new Error('No AI provider available');
-
-  const systemMsg = messages.find(m => m.role === 'system');
-  const userMessages = messages.filter(m => m.role !== 'system');
-
-  const model = ANTHROPIC_PRIMARY_MODEL;
-  const payload: any = {
-    model,
-    max_tokens: options.maxTokens ?? 2000,
-    stream: true,
-    messages: userMessages.map(m => ({ role: m.role, content: m.content })),
-  };
-  if (systemMsg) payload.system = systemMsg.content;
-
-  let response = await fetch('https://api.anthropic.com/v1/messages', {
-    method: 'POST',
-    headers: {
-      'Content-Type': 'application/json',
-      'x-api-key': apiKey,
-      'anthropic-version': '2023-06-01',
-    },
-    body: JSON.stringify(payload),
-  });
-
-  if (!response.ok) {
-    // Try fallback model
-    if (response.status === 404) {
-      const fallbackPayload = { ...payload, model: ANTHROPIC_FALLBACK_MODEL };
-      const fallbackResponse = await fetch('https://api.anthropic.com/v1/messages', {
-        method: 'POST',
-        headers: {
-          'Content-Type': 'application/json',
-          'x-api-key': apiKey,
-          'anthropic-version': '2023-06-01',
-        },
-        body: JSON.stringify(fallbackPayload),
-      });
-      if (!fallbackResponse.ok || !fallbackResponse.body) throw new Error('Anthropic unavailable');
-      logHealthEvent({ provider: 'anthropic', success: true, latencyMs: 0, model: ANTHROPIC_FALLBACK_MODEL, priority, isFailover: true });
-      return { stream: fallbackResponse.body, provider: 'anthropic', model: ANTHROPIC_FALLBACK_MODEL };
-    }
-    throw new Error(`Anthropic HTTP ${response.status}`);
-  }
-
-  if (!response.body) throw new Error('No response body from Anthropic');
-  logHealthEvent({ provider: 'anthropic', success: true, latencyMs: 0, model, priority, isFailover: !!endpoint });
-  return { stream: response.body, provider: 'anthropic', model };
+  // No-paid-AI policy: no Anthropic fallback. If Ollama streaming failed,
+  // surface that clearly so the caller can degrade gracefully.
+  throw new Error('Ollama streaming unavailable — no paid-AI fallback per policy');
 }
 
 // ─── Ollama model list ────────────────────────────────────────────────────────