Move ingester telemetry setup behind argument parsing

CHANGELOG.md

@@ -4,7 +4,7 @@
 ### Fixed
 
 - `cross-encoder` reranking no longer ties the scores of strongly-relevant candidates, which left their order to the sort. Scores remain 0-1.
-- haiku-rag CLI startup no longer imports `lancedb`, `pyarrow` and `pydantic_ai`.
+- `haiku-rag` and `haiku-ingester` CLI startup no longer imports `lancedb`, `pyarrow` and `pydantic_ai`.
 - `haiku.rag.store` no longer re-exports `Store`; import it from `haiku.rag.store.engine`.
 
 ## [0.73.0] - 2026-08-06

haiku_rag_slim/haiku/rag/ingester/cli.py

@@ -59,15 +59,15 @@ def main(
 ) -> None:
     """Top-level callback so every subcommand inherits --config without
     each one redeclaring it. Mirrors haiku-rag's CLI shape."""
+    from haiku.rag.telemetry import configure as configure_telemetry
+
     _load_config_with_override(config)
+    configure_cli_logging()
+    configure_telemetry(service_name="haiku-ingester")
 
 
 def cli() -> None:
     """Entry point that translates store-state errors into a clean exit."""
-    from haiku.rag.telemetry import configure as configure_telemetry
-
-    configure_cli_logging()
-    configure_telemetry(service_name="haiku-ingester")
     try:
         _cli()
     except (MigrationRequiredError, ReadOnlyError) as e:

tests/ingester/test_cli.py

@@ -26,15 +26,17 @@ from haiku.rag.ingester.queue.models import JobOp
 runner = CliRunner()
 
 
-def test_importing_ingester_cli_does_not_load_lancedb():
+def test_importing_ingester_cli_does_not_load_heavy_dependencies():
-    """Test that lancedb is not imported automatically by the cli. Doing so is
+    """Importing the CLI must not pull the heavy runtime dependencies; they cost
-     expensive. Must be run in a subprocess because lancedb might be imported by
+    seconds of startup. Runs in a subprocess because another test in the same
-    other tests in the same session."""
+    session may already have imported them."""
     result = subprocess.run(
         [
             sys.executable,
             "-c",
-            "import haiku.rag.ingester.cli, sys; assert 'lancedb' not in sys.modules",
+            "import haiku.rag.ingester.cli, sys; "
+            "loaded = {'lancedb', 'pyarrow', 'pydantic_ai'} & sys.modules.keys(); "
+            "assert not loaded, loaded",
         ],
         capture_output=True,
         text=True,

tests/test_cli.py

@@ -14,15 +14,17 @@ from haiku.rag.store.exceptions import MigrationRequiredError
 runner = CliRunner()
 
 
-def test_importing_cli_does_not_load_lancedb():
+def test_importing_cli_does_not_load_heavy_dependencies():
-    """Test that lancedb is not imported automatically by the cli. Doing so is
+    """Importing the CLI must not pull the heavy runtime dependencies; they cost
-     expensive. Must be run in a subprocess because lancedb might be imported by
+    seconds of startup. Runs in a subprocess because another test in the same
-    other tests in the same session."""
+    session may already have imported them."""
     result = subprocess.run(
         [
             sys.executable,
             "-c",
-            "import haiku.rag.cli, sys; assert 'lancedb' not in sys.modules",
+            "import haiku.rag.cli, sys; "
+            "loaded = {'lancedb', 'pyarrow', 'pydantic_ai'} & sys.modules.keys(); "
+            "assert not loaded, loaded",
         ],
         capture_output=True,
         text=True,