fix: defer tantivy/textual imports to avoid SIGSEGV on Linux CI

Tantivy (Rust) was crashing with SIGSEGV when loaded at CLI startup on
ubuntu-latest. Defer imports to explore command only so backup/sync
work without loading native search deps.

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: pascalwhoop

.github/workflows/release.yml

@@ -37,7 +37,14 @@ jobs:
         run: uv sync
 
       - name: Run tests
-        run: uv run pytest -v --tb=short
+        run: uv run pytest -v --tb=short 2>&1 | tee test-output.txt; exit ${PIPESTATUS[0]}
+
+      - name: Upload test output on failure
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: test-output
+          path: test-output.txt
 
   version:
     name: Bump version and create tag

src/convx_ai/cli.py

@@ -13,8 +13,6 @@
 
 from convx_ai.adapters import default_input_path, get_adapter
 from convx_ai.engine import SyncResult, sync_sessions
-from convx_ai.search import ensure_index
-from convx_ai.tui import ExploreApp
 from convx_ai.utils import sanitize_segment
 
 app = typer.Typer(help="Export AI conversations into a Git repo.", no_args_is_help=True)
@@ -233,6 +231,9 @@ def explore_command(
     ),
 ) -> None:
     """Browse and search exported conversations in a TUI."""
+    from convx_ai.search import ensure_index
+    from convx_ai.tui import ExploreApp
+
     repo = _require_git_repo(output_path)
     index_path = repo / ".convx" / "index.json"
     if not index_path.exists():

src/convx_ai/search.py

@@ -4,10 +4,10 @@
 import shutil
 from pathlib import Path
 
-from tantivy import Document, Index, SchemaBuilder
-
 
 def _schema():
+    from tantivy import SchemaBuilder
+
     return (
         SchemaBuilder()
         .add_text_field("session_key", stored=True)
@@ -34,6 +34,8 @@ def ensure_index(repo: Path) -> None:
     if search_dir.exists():
         shutil.rmtree(search_dir)
     search_dir.mkdir(parents=True)
+    from tantivy import Document, Index
+
     schema = _schema()
     index = Index(schema=schema, path=str(search_dir))
     writer = index.writer(heap_size=15_000_000, num_threads=1)
@@ -84,6 +86,8 @@ def list_sessions(repo: Path) -> list[dict]:
 
 
 def query_index(repo: Path, q: str, limit: int = 50) -> list[dict]:
+    from tantivy import Index
+
     search_dir = repo / ".convx" / "search-index"
     if not search_dir.exists():
         return []