fix(ollama): disable thinking for deepseek call sites

apps/api/src/hermes/nl_gateway.py

@@ -276,6 +276,8 @@ async def process_nl_message(
                     f"{endpoint.url}/api/chat",
                     json={
                         "model": model,
+                        # Keep Hermes responses in message.content across Ollama 0.24+.
+                        "think": False,
                         "messages": [
                             {"role": "system", "content": system_prompt},
                             {"role": "user",   "content": prompt_with_ctx},