diff --git a/navi/core/registry.py b/navi/core/registry.py
index 4cb3940..99f6820 100644
--- a/navi/core/registry.py
+++ b/navi/core/registry.py
@@ -116,6 +116,11 @@
     from navi.llm.ollama import OllamaBackend
     from navi.llm.fallback import FallbackOllamaBackend
     from navi.llm.openai_backend import OpenAIBackend
+    ollama_http_timeout = max(
+        settings.ollama_request_timeout,
+        settings.llm_complete_timeout,
+        settings.llm_stream_first_chunk_timeout,
+    )
 
     # Ollama backend (primary)
     if settings.ollama_backends_file:
@@ -126,6 +131,7 @@
             model=settings.ollama_default_model,
             host=settings.ollama_host,
             api_key=settings.ollama_api_key,
+            timeout=ollama_http_timeout,
         )))
 
     # OpenAI backend (if configured)
diff --git a/navi/llm/fallback.py b/navi/llm/fallback.py
index 163d87f..9bb8106 100644
--- a/navi/llm/fallback.py
+++ b/navi/llm/fallback.py
@@ -79,9 +79,14 @@
     def _get_client(self, server: ServerEntry) -> OllamaBackend:
         from navi.config import settings
         if server.host not in self._clients:
+            ollama_http_timeout = max(
+                settings.ollama_request_timeout,
+                settings.llm_complete_timeout,
+                settings.llm_stream_first_chunk_timeout,
+            )
             self._clients[server.host] = OllamaBackend(
                 model="", host=server.host, api_key=server.api_key,
-                timeout=settings.ollama_request_timeout,
+                timeout=ollama_http_timeout,
             )
         return self._clients[server.host]
 
diff --git a/navi/llm/ollama.py b/navi/llm/ollama.py
index aa5db4c..5e0146c 100644
--- a/navi/llm/ollama.py
+++ b/navi/llm/ollama.py
@@ -71,7 +71,10 @@
             return LLMModelNotFoundError(str(e))
         return LLMBackendError(str(e))
     # Catch httpx / socket connection failures by message
+    err_type = type(e).__name__.lower()
     err_str = str(e).lower()
+    if "timeout" in err_type:
+        return LLMConnectionError(str(e) or type(e).__name__)
     if any(kw in err_str for kw in ("connect", "connection refused", "name or service not known",
                                      "network", "timeout", "unreachable", "nodename")):
         return LLMConnectionError(str(e))
diff --git a/tests/unit/llm/test_ollama.py b/tests/unit/llm/test_ollama.py
new file mode 100644
index 0000000..10d6b00
--- /dev/null
+++ b/tests/unit/llm/test_ollama.py
@@ -0,0 +1,19 @@
+"""Unit tests for Ollama backend helpers."""
+
+import httpx
+
+from navi.exceptions import LLMConnectionError
+from navi.llm.ollama import _classify_error
+
+
+def test_classify_read_timeout_as_connection_error():
+    err = _classify_error(httpx.ReadTimeout("timed out"))
+
+    assert isinstance(err, LLMConnectionError)
+
+
+def test_classify_empty_timeout_message_as_connection_error():
+    err = _classify_error(httpx.ReadTimeout(""))
+
+    assert isinstance(err, LLMConnectionError)
+    assert str(err) == "ReadTimeout"