fix: model introspection uses exact match, queries /api/ps first

_get_ollama_model() used prefix match (startswith) on /api/tags,
causing qwen3:30b to match qwen3.5:latest. Now:
1. Queries /api/ps (loaded models) first — most accurate
2. Falls back to /api/tags with exact name match
3. Reports actual running model, not just configured one

Updated test_get_system_info_contains_model to not assume model==config.

Fixes #77. 5 regression tests added.

tests/timmy/test_introspection.py

@@ -1,5 +1,9 @@
 """Tests for system introspection tools."""
 
+from unittest.mock import MagicMock, patch
+
+import httpx
+
 
 def test_get_system_info_returns_dict():
     """System info should return a dictionary."""
@@ -15,15 +19,17 @@ def test_get_system_info_returns_dict():
 
 
 def test_get_system_info_contains_model():
-    """System info should include model name."""
-    from config import settings
+    """System info should include a model name (may differ from config if
+    the actual running model is different — see issue #77)."""
     from timmy.tools_intro import get_system_info
 
     info = get_system_info()
 
     assert "model" in info
-    # Model should come from settings
-    assert info["model"] == settings.ollama_model
+    # Model should be a non-empty string — exact value depends on what
+    # Ollama has loaded (verified by TestGetOllamaModelExactMatch tests)
+    assert isinstance(info["model"], str)
+    assert len(info["model"]) > 0
 
 
 def test_get_system_info_contains_repo_root():
@@ -59,3 +65,96 @@ def test_get_memory_status_returns_dict():
     assert isinstance(status, dict)
     assert "tier1_hot_memory" in status
     assert "tier2_vault" in status
+
+
+# --- _get_ollama_model exact-match tests (issue #77) ---
+
+
+def _mock_response(json_data, status_code=200):
+    """Create a mock httpx response."""
+    resp = MagicMock(spec=httpx.Response)
+    resp.status_code = status_code
+    resp.json.return_value = json_data
+    return resp
+
+
+class TestGetOllamaModelExactMatch:
+    """Ensure _get_ollama_model uses exact match, not prefix match."""
+
+    @patch("timmy.tools_intro.httpx.get")
+    def test_exact_match_from_ps(self, mock_get):
+        """Should return exact model from /api/ps."""
+        from timmy.tools_intro import _get_ollama_model
+
+        ps_resp = _mock_response({"models": [{"name": "qwen3:30b"}]})
+        mock_get.return_value = ps_resp
+
+        with patch("config.settings") as mock_settings:
+            mock_settings.ollama_model = "qwen3:30b"
+            mock_settings.ollama_url = "http://localhost:11434"
+            result = _get_ollama_model()
+
+        assert result == "qwen3:30b"
+
+    @patch("timmy.tools_intro.httpx.get")
+    def test_prefix_collision_returns_correct_model(self, mock_get):
+        """qwen3:30b configured — must NOT match qwen3.5:latest (prefix bug)."""
+        from timmy.tools_intro import _get_ollama_model
+
+        # /api/ps has both models loaded; configured is qwen3:30b
+        ps_resp = _mock_response({"models": [{"name": "qwen3.5:latest"}, {"name": "qwen3:30b"}]})
+        mock_get.return_value = ps_resp
+
+        with patch("config.settings") as mock_settings:
+            mock_settings.ollama_model = "qwen3:30b"
+            mock_settings.ollama_url = "http://localhost:11434"
+            result = _get_ollama_model()
+
+        assert result == "qwen3:30b", f"Got '{result}' — prefix collision bug!"
+
+    @patch("timmy.tools_intro.httpx.get")
+    def test_configured_model_not_running_returns_actual(self, mock_get):
+        """If configured model isn't loaded, report what IS running."""
+        from timmy.tools_intro import _get_ollama_model
+
+        ps_resp = _mock_response({"models": [{"name": "qwen3.5:latest"}]})
+        mock_get.return_value = ps_resp
+
+        with patch("config.settings") as mock_settings:
+            mock_settings.ollama_model = "qwen3:30b"
+            mock_settings.ollama_url = "http://localhost:11434"
+            result = _get_ollama_model()
+
+        # Should report actual running model, not configured one
+        assert result == "qwen3.5:latest"
+
+    @patch("timmy.tools_intro.httpx.get")
+    def test_latest_suffix_match(self, mock_get):
+        """'qwen3:30b' config should match 'qwen3:30b:latest' from API."""
+        from timmy.tools_intro import _get_ollama_model
+
+        ps_resp = _mock_response({"models": []})
+        tags_resp = _mock_response({"models": [{"name": "qwen3:30b:latest"}]})
+        mock_get.side_effect = [ps_resp, tags_resp]
+
+        with patch("config.settings") as mock_settings:
+            mock_settings.ollama_model = "qwen3:30b"
+            mock_settings.ollama_url = "http://localhost:11434"
+            result = _get_ollama_model()
+
+        # Falls back to configured since no exact match
+        assert result == "qwen3:30b"
+
+    @patch("timmy.tools_intro.httpx.get")
+    def test_ollama_down_returns_configured(self, mock_get):
+        """If Ollama is unreachable, return configured model."""
+        from timmy.tools_intro import _get_ollama_model
+
+        mock_get.side_effect = httpx.ConnectError("connection refused")
+
+        with patch("config.settings") as mock_settings:
+            mock_settings.ollama_model = "qwen3:30b"
+            mock_settings.ollama_url = "http://localhost:11434"
+            result = _get_ollama_model()
+
+        assert result == "qwen3:30b"