diff --git a/tests/test_node_execution_contract.py b/tests/test_node_execution_contract.py
index 05d3be0..a75a423 100644
--- a/tests/test_node_execution_contract.py
+++ b/tests/test_node_execution_contract.py
@@ -138,5 +138,291 @@ class TestIsChangedPromptSearchList(unittest.TestCase):
         self.assertIsInstance(result, float)
 
 
+import logging  # noqa: E402
+import tempfile  # noqa: E402
+import types  # noqa: E402
+import unittest.mock as mock  # noqa: E402
+
+from database.operations import PromptDatabase  # noqa: E402
+from prompt_manager_base import PromptManagerBase  # noqa: E402
+import prompt_manager_base  # noqa: E402
+
+
+class FakeClip:
+    """The two CLIP calls the node makes, recorded."""
+
+    def __init__(self):
+        self.tokenized = []
+
+    def tokenize(self, text):
+        self.tokenized.append(text)
+        return ("tokens", text)
+
+    def encode_from_tokens_scheduled(self, tokens):
+        return [("cond", tokens[1])]
+
+
+class NodeTestCase(unittest.TestCase):
+    """A node with a real temporary database and no monitor or tracker."""
+
+    node_cls = PromptManagerBase
+
+    def setUp(self):
+        fd, self.db_path = tempfile.mkstemp(suffix=".db")
+        os.close(fd)
+        self.db = PromptDatabase(self.db_path)
+        self.node = self.node_cls.__new__(self.node_cls)
+        self.node.db = self.db
+        self.node.logger = logging.getLogger("test.node")
+        self.node.prompt_tracker = mock.Mock()
+        self.node.image_monitor = mock.Mock()
+        self.node.comfyui_integration = mock.Mock()
+
+    def tearDown(self):
+        for suffix in ("", "-wal", "-shm"):
+            if os.path.exists(self.db_path + suffix):
+                os.unlink(self.db_path + suffix)
+
+
+class TestBaseSaveAndSearch(NodeTestCase):
+    def test_init_wires_database_tracker_monitor_and_integration(self):
+        with (
+            mock.patch.object(prompt_manager_base, "PromptDatabase") as db_cls,
+            mock.patch.object(prompt_manager_base, "get_prompt_tracker") as tracker,
+            mock.patch.object(prompt_manager_base, "get_image_monitor") as monitor,
+            mock.patch.object(
+                prompt_manager_base, "get_comfyui_integration"
+            ) as integration,
+        ):
+            node = PromptManagerBase()
+        self.assertIs(node.db, db_cls.return_value)
+        tracker.assert_called_once_with(db_cls.return_value)
+        monitor.assert_called_once_with(db_cls.return_value, tracker.return_value)
+        self.assertIs(node.comfyui_integration, integration.return_value)
+        monitor.return_value.start_monitoring.assert_called_once_with()
+
+    def test_saving_an_existing_prompt_updates_its_metadata(self):
+        first = self.node._save_prompt_to_database("a cat")
+        second = self.node._save_prompt_to_database(
+            "a cat", category="animals", tags=["cute"]
+        )
+        self.assertEqual(first, second)
+        saved = self.db.get_prompt_by_id(first)
+        self.assertEqual(saved["category"], "animals")
+        self.assertIn("cute", saved["tags"])
+        self.assertEqual(saved["run_count"], 0)
+
+    def test_count_run_on_an_existing_prompt(self):
+        prompt_id = self.node._save_prompt_to_database("a cat")
+        self.node._save_prompt_to_database("a cat", count_run=True)
+        self.assertEqual(self.db.get_prompt_by_id(prompt_id)["run_count"], 1)
+
+    def test_database_errors_are_logged_and_return_none(self):
+        self.node.db = mock.Mock()
+        self.node.db.get_prompt_by_hash.side_effect = RuntimeError("locked")
+        with self.assertLogs("test.node", level="ERROR"):
+            self.assertIsNone(self.node._save_prompt_to_database("a cat"))
+
+    def test_record_use_failure_is_a_warning(self):
+        self.node.db = mock.Mock()
+        self.node.db.record_prompt_use.side_effect = RuntimeError("locked")
+        with self.assertLogs("test.node", level="WARNING"):
+            self.node._record_use(1)
+
+    def test_parse_tags(self):
+        self.assertIsNone(self.node._parse_tags(""))
+        self.assertIsNone(self.node._parse_tags(" , ,"))
+        self.assertEqual(self.node._parse_tags(" a, b ,,c"), ["a", "b", "c"])
+
+    def test_search_prompts_and_api_wrappers(self):
+        self.node._save_prompt_to_database("a cat on a mat")
+        self.assertEqual(self.node._search_prompts(""), [])
+        self.assertEqual(self.node._search_prompts("   "), [])
+        self.assertEqual(len(self.node.search_prompts_api("cat")), 1)
+        self.assertEqual(len(self.node.get_recent_prompts_api(limit=5)["prompts"]), 1)
+
+    def test_search_and_recent_errors_return_empty_lists(self):
+        self.node.db = mock.Mock()
+        self.node.db.search_prompts.side_effect = RuntimeError("locked")
+        self.node.db.get_recent_prompts.side_effect = RuntimeError("locked")
+        with self.assertLogs("test.node", level="ERROR"):
+            self.assertEqual(self.node._search_prompts("cat"), [])
+        with self.assertLogs("test.node", level="ERROR"):
+            self.assertEqual(self.node.get_recent_prompts_api(), [])
+
+    def test_gallery_system_start_failure_is_logged(self):
+        self.node.image_monitor.start_monitoring.side_effect = OSError("no inotify")
+        with self.assertLogs("test.node", level="ERROR"):
+            self.node._start_gallery_system()
+        self.node.image_monitor.get_status.return_value = {"running": False}
+        self.node.prompt_tracker.get_status.return_value = {"active_prompts_count": 0}
+        self.assertEqual(
+            self.node.get_gallery_status(),
+            {
+                "image_monitor": {"running": False},
+                "prompt_tracker": {"active_prompts_count": 0},
+            },
+        )
+        self.node.cleanup_gallery_system()
+        self.node.image_monitor.stop_monitoring.assert_not_called()
+
+
+def fake_integration_config(enabled, trigger_words=True, path="/loras"):
+    module = types.ModuleType("py.config")
+    module.IntegrationConfig = types.SimpleNamespace(
+        LORA_MANAGER_ENABLED=enabled,
+        LORA_TRIGGER_WORDS_ENABLED=trigger_words,
+        LORA_MANAGER_PATH=path,
+    )
+    return module
+
+
+class TestLoraTriggerWords(NodeTestCase):
+    def test_disabled_integration_returns_text_unchanged(self):
+        with mock.patch.dict(
+            sys.modules, {"py.config": fake_integration_config(False)}
+        ):
+            self.assertEqual(self.node._inject_lora_trigger_words("a cat"), "a cat")
+        with mock.patch.dict(
+            sys.modules, {"py.config": fake_integration_config(True, False)}
+        ):
+            self.assertEqual(self.node._inject_lora_trigger_words("a cat"), "a cat")
+
+    def test_missing_config_or_lora_utils_returns_text_unchanged(self):
+        with mock.patch.dict(sys.modules, {"py.config": None}):
+            self.assertEqual(self.node._inject_lora_trigger_words("a cat"), "a cat")
+        with mock.patch.dict(
+            sys.modules,
+            {"py.config": fake_integration_config(True), "py.lora_utils": None},
+        ):
+            self.assertEqual(self.node._inject_lora_trigger_words("a cat"), "a cat")
+
+    def test_enabled_integration_loads_the_cache_and_injects(self):
+        cache = mock.Mock(is_loaded=False)
+        lora_utils = types.ModuleType("py.lora_utils")
+        lora_utils.get_trigger_cache = lambda: cache
+        lora_utils.inject_trigger_words = lambda text, c: (text + " trig", ["trig"])
+        with mock.patch.dict(
+            sys.modules,
+            {"py.config": fake_integration_config(True), "py.lora_utils": lora_utils},
+        ):
+            self.assertEqual(
+                self.node._inject_lora_trigger_words("a cat"), "a cat trig"
+            )
+        cache.load.assert_called_once_with("/loras")
+
+
+class TestPromptManagerEncoding(NodeTestCase):
+    node_cls = PromptManager
+
+    def setUp(self):
+        super().setUp()
+        self.node._inject_lora_trigger_words = lambda text: text
+
+    def test_missing_clip_is_a_clear_error(self):
+        with self.assertRaisesRegex(RuntimeError, "clip input is invalid"):
+            self.node.encode_prompt(None, "a cat")
+
+    def test_encodes_combined_text_and_saves_the_main_text(self):
+        clip = FakeClip()
+        conditioning, final = self.node.encode_prompt(
+            clip, "a cat", prepend_text=" best ", append_text="8k ", tags="x"
+        )
+        self.assertEqual(final, "best a cat 8k")
+        self.assertEqual(clip.tokenized, ["best a cat 8k"])
+        self.assertEqual(conditioning, [("cond", "best a cat 8k")])
+        saved = self.db.get_recent_prompts(limit=1)["prompts"][0]
+        self.assertEqual(saved["text"], "a cat")
+        self.assertIn("prepend:best", saved["tags"])
+        self.assertIn("append:8k", saved["tags"])
+        self.node.comfyui_integration.register_prompt.assert_called_once()
+
+    def test_tracking_failure_does_not_break_encoding(self):
+        self.node._track_prompt_execution = mock.Mock(side_effect=RuntimeError("db"))
+        with self.assertLogs("test.node", level="WARNING"):
+            _, final = self.node.encode_prompt(FakeClip(), "a cat")
+        self.assertEqual(final, "a cat")
+
+    def test_empty_text_is_encoded_but_not_saved(self):
+        _, final = self.node.encode_prompt(FakeClip(), "   ")
+        self.assertEqual(final.strip(), "")
+        self.assertEqual(self.db.get_recent_prompts(limit=1)["prompts"], [])
+
+
+class TestPromptManagerTextProcessing(NodeTestCase):
+    node_cls = PromptManagerText
+
+    def setUp(self):
+        super().setUp()
+        self.node._inject_lora_trigger_words = lambda text: text
+
+    def test_tracking_failure_does_not_break_processing(self):
+        self.node._track_prompt_execution = mock.Mock(side_effect=RuntimeError("db"))
+        with self.assertLogs("test.node", level="WARNING"):
+            result = self.node.process_text("a castle", prepend_text="big")
+        self.assertEqual(result[0], "big a castle")
+
+
+class TestPromptSearchListSearch(NodeTestCase):
+    node_cls = PromptSearchList
+
+    def _save(self, text, **kwargs):
+        from utils.hashing import generate_prompt_hash
+
+        return self.db.save_prompt(
+            text=text, prompt_hash=generate_prompt_hash(text), **kwargs
+        )
+
+    def test_init_opens_the_database(self):
+        with mock.patch("prompt_search_list.PromptDatabase") as db_cls:
+            node = PromptSearchList()
+        self.assertIs(node.db, db_cls.return_value)
+
+    def test_no_results_returns_a_single_empty_item(self):
+        result = self.node.search(text="nothing here")
+        self.assertEqual(result["result"], ([""], "No results found", 0))
+        self.assertEqual(result["ui"], {"text": ["No results found"]})
+
+    def test_filters_by_text_category_tags_and_rating(self):
+        self._save("a red cat", category="animals", tags=["cute"])
+        self._save("a blue car", category="vehicles", tags=["fast"])
+        self.assertEqual(self.node.search(text="cat")["result"][0], ["a red cat"])
+        self.assertEqual(
+            self.node.search(category="vehicles")["result"][0], ["a blue car"]
+        )
+        self.assertEqual(self.node.search(tags=" cute , ")["result"][0], ["a red cat"])
+        self.assertEqual(self.node.search(min_rating=5)["result"][2], 0)
+        self.assertEqual(self.node.search(limit=1)["result"][2], 1)
+
+    def test_multipart_and_lora_only_prompts_are_skipped(self):
+        self._save("Clip_1: a cat Clip_2: a dog")
+        self._save("<lora:style:0.8> <lora:other:1>")
+        self._save("plain prompt")
+        self.assertEqual(self.node.search()["result"][0], ["plain prompt"])
+        result = self.node.search(skip_multipart=False)["result"]
+        self.assertEqual(
+            sorted(result[0]), ["Clip_1: a cat Clip_2: a dog", "plain prompt"]
+        )
+
+    def test_newlines_collapse_and_preview_is_numbered_and_truncated(self):
+        self._save("line one\n   line two")
+        self._save("x" * 200)
+        result = self.node.search()
+        prompts, preview, count = result["result"]
+        self.assertEqual(count, 2)
+        self.assertIn("line one line two", prompts)
+        lines = preview.splitlines()
+        self.assertTrue(lines[0].startswith("[1] "))
+        self.assertTrue(any(line.endswith("...") and len(line) < 140 for line in lines))
+        self.assertEqual(result["ui"], {"text": ["Found 2 prompts"]})
+
+    def test_database_errors_are_returned_not_raised(self):
+        self.node.db = mock.Mock()
+        self.node.db.search_prompts.side_effect = RuntimeError("locked")
+        result = self.node.search(text="cat")
+        self.assertEqual(result["result"], ([""], "Error: locked", 0))
+        self.assertEqual(result["ui"], {"text": ["Search error: locked"]})
+
+
 if __name__ == "__main__":
     unittest.main()