Merge pull request #9 from M9nx/copilot/add-batch-size-option

Add configurable embedding batch size to indexing CLI

Author: M9nx

semantic_code_intelligence/cli/commands/index_cmd.py

@@ -10,7 +10,7 @@
 
 import click
 
-from semantic_code_intelligence.config.settings import AppConfig
+from semantic_code_intelligence.config.settings import AppConfig, load_config, save_config
 from semantic_code_intelligence.services.indexing_service import run_indexing
 from semantic_code_intelligence.utils.logging import (
     get_logger,
@@ -201,7 +201,14 @@ def _on_changes(events: list) -> None:
     type=str,
     help="Switch embedding model and re-index in one step.",
 )
-def index_cmd(project_path: Path | None, force: bool, watch: bool, add_file: str | None, inspect_file: str | None, switch_model: str | None) -> int:
+@click.option(
+    "--batch-size",
+    "batch_size",
+    type=click.IntRange(1, None),
+    default=None,
+    help="Embedding batch size for chunk processing (overrides config).",
+)
+def index_cmd(project_path: Path | None, force: bool, watch: bool, add_file: str | None, inspect_file: str | None, switch_model: str | None, batch_size: int | None) -> int:
     """Index a codebase for semantic search.
 
     Scans the target directory, extracts code chunks, generates embeddings,
@@ -229,6 +236,8 @@ def index_cmd(project_path: Path | None, force: bool, watch: bool, add_file: str
             f"Project not initialized at {root}. Run 'codexa init' first."
         )
 
+    config: AppConfig | None = None
+
     # --- Inspect mode: show metadata for a file ---
     if inspect_file:
         _inspect_file_index(root, inspect_file)
@@ -242,10 +251,9 @@ def index_cmd(project_path: Path | None, force: bool, watch: bool, add_file: str
     # --- Switch model inline: update config + force re-index ---
     if switch_model:
         from semantic_code_intelligence.embeddings.model_registry import resolve_model_name
-        from semantic_code_intelligence.config.settings import load_config, save_config
 
         resolved = resolve_model_name(switch_model)
-        config = load_config(root)
+        config = config if config is not None else load_config(root)
         old_model = config.embedding.model_name
         if old_model == resolved:
             print_info(f"Model already set to '{resolved}' — running normal index.")
@@ -255,10 +263,23 @@ def index_cmd(project_path: Path | None, force: bool, watch: bool, add_file: str
             print_success(f"Switched model: {old_model} → {resolved}")
         force = True  # force re-index with new model
 
+    # --- Optional batch size override (only when indexing will run) ---
+    if batch_size is not None:
+        config = config if config is not None else load_config(root)
+        prev_batch = config.embedding.batch_size
+        if prev_batch != batch_size:
+            config.embedding.batch_size = batch_size
+            save_config(config, root)
+            print_info(
+                f"Embedding batch size updated: {prev_batch} → {batch_size} "
+                "(applies to this and future indexing runs)."
+            )
+        else:
+            print_info(f"Embedding batch size already set to {batch_size}.")
+
     # --- Model consistency guard ---
     if not force:
         from semantic_code_intelligence.storage.index_manifest import IndexManifest
-        from semantic_code_intelligence.config.settings import load_config
         index_dir = AppConfig.index_dir(root)
         manifest = IndexManifest.load(index_dir)
         if manifest:

semantic_code_intelligence/tests/test_cli.py

@@ -8,6 +8,7 @@
 
 import pytest
 from click.testing import CliRunner
+from semantic_code_intelligence.config.settings import load_config
 
 from semantic_code_intelligence.cli.main import cli
 from semantic_code_intelligence.embeddings.generator import BYTES_PER_GB
@@ -133,6 +134,37 @@ def test_index_force_flag(self, runner: CliRunner, tmp_path: Path):
         result = runner.invoke(cli, ["index", str(tmp_path), "--force"])
         assert result.exit_code == 0
 
+    def test_index_batch_size_override(self, runner: CliRunner, tmp_path: Path, monkeypatch: pytest.MonkeyPatch):
+        project = tmp_path
+        (project / "sample.py").write_text("def foo():\n    return 1\n", encoding="utf-8")
+
+        captured: dict[str, int] = {}
+
+        class DummyResult:
+            files_scanned = 1
+            files_indexed = 1
+            files_skipped = 0
+            chunks_created = 1
+            total_vectors = 1
+
+        def fake_run_indexing(project_root, force=False):
+            cfg = load_config(project_root)
+            captured["batch_size"] = cfg.embedding.batch_size
+            return DummyResult()
+
+        monkeypatch.setattr(
+            "semantic_code_intelligence.cli.commands.index_cmd.run_indexing",
+            fake_run_indexing,
+        )
+
+        runner.invoke(cli, ["init", str(project)])
+        result = runner.invoke(cli, ["index", str(project), "--batch-size", "8"])
+        assert result.exit_code == 0
+
+        config = json.loads((project / ".codexa" / "config.json").read_text(encoding="utf-8"))
+        assert config["embedding"]["batch_size"] == 8
+        assert captured["batch_size"] == 8
+
     def test_index_network_oserror_is_nonfatal(self, runner: CliRunner, tmp_path: Path, monkeypatch: pytest.MonkeyPatch):
         runner.invoke(cli, ["init", str(tmp_path)])