fix: reconcile run_vector_search defaults with CLI defaults

src/vectrify/tests/vector/test_runner.py

@@ -1,13 +1,28 @@
+import inspect
+
 import pytest
 from PIL import Image
 
+from vectrify import cli
 from vectrify.formats.svg.plugin import SvgPlugin
 from vectrify.score import ScorerType
 from vectrify.search import StrategyType
 from vectrify.vector.runner import run_vector_search
 from vectrify.vector.storage import FileStorageAdapter
 
 
+def test_runner_defaults_match_cli_defaults():
+    defaults = {
+        p.name: p.default
+        for p in inspect.signature(run_vector_search).parameters.values()
+    }
+    assert defaults["pool_size"] == cli.DEFAULT_POOL_SIZE
+    assert defaults["epoch_diversity"] == cli.DEFAULT_EPOCH_DIVERSITY
+    assert defaults["vision_model"] == cli.DEFAULT_VISION_MODEL
+    # llm_rate has no static default; it is derived from --workers at call time.
+    assert defaults["llm_rate"] is None
+
+
 @pytest.mark.llm
 def test_run_svg_search_end_to_end(tmp_path):
     img_path = tmp_path / "test.png"

src/vectrify/vector/runner.py

@@ -12,6 +12,12 @@
 
 from PIL import Image
 
+from vectrify.cli import (
+    DEFAULT_EPOCH_DIVERSITY,
+    DEFAULT_POOL_SIZE,
+    DEFAULT_VISION_MODEL,
+    _default_llm_rate,
+)
 from vectrify.formats.models import VectorStatePayload
 from vectrify.image_utils import (
     downscale_png_bytes,
@@ -121,21 +127,24 @@ def run_vector_search(
     save_raster: bool = False,
     epoch_patience: int | None = None,
     epoch_min_delta: float = 1e-4,
-    llm_rate: float = 0.2,
-    pool_size: int = 20,
+    llm_rate: float | None = None,
+    pool_size: int = DEFAULT_POOL_SIZE,
     seeds: int = 0,
     beams: int = 10,
     cull_keep: float = 0.5,
-    epoch_diversity: float = 0.10,
+    epoch_diversity: float = DEFAULT_EPOCH_DIVERSITY,
     epoch_variance: float | None = None,
     max_epochs: int | None = None,
     epoch_pool_size: int | None = None,
     epoch_steps: int | None = None,
     max_llm_calls: int | None = None,
-    vision_model: str = "ensemble",
+    vision_model: str = DEFAULT_VISION_MODEL,
     stats: "SearchStats | None" = None,
     dashboard: "Dashboard | None" = None,
 ) -> None:
+    if llm_rate is None:
+        llm_rate = _default_llm_rate(workers)
+
     storage.initialize()
     assert storage.current_run_dir is not None
     run_log_file = storage.current_run_dir / "search.log"