fix(ingest): make memory-doc filenames unique so concurrent saves do not overwrite

graphify/ingest.py

@@ -2,6 +2,7 @@
 from __future__ import annotations
 import json
 import re
+import uuid
 import urllib.error
 import urllib.parse
 from datetime import datetime, timezone
@@ -299,7 +300,11 @@ def save_query_result(
 
     now = datetime.now(timezone.utc)
     slug = re.sub(r"[^\w]", "_", question.lower())[:50].strip("_")
-    filename = f"query_{now.strftime('%Y%m%d_%H%M%S')}_{slug}.md"
+    # A second-granularity stamp plus a 50-char slug is not unique: two saves in
+    # the same second whose questions share a prefix resolve to one path, and the
+    # later write_text silently replaces the earlier one (#3301). The short uuid
+    # makes every save its own file; the query_ prefix and .md suffix are kept.
+    filename = f"query_{now.strftime('%Y%m%d_%H%M%S')}_{uuid.uuid4().hex[:8]}_{slug}.md"
 
     frontmatter_lines = [
         "---",

tests/test_ingest.py

@@ -98,3 +98,15 @@ def test_no_outcome_means_no_outcome_section(tmp_path):
 def test_invalid_outcome_rejected(tmp_path):
     with pytest.raises(ValueError):
         save_query_result("q", "a", tmp_path / "memory", outcome="great")
+
+
+def test_concurrent_saves_of_the_same_question_do_not_overwrite(tmp_path):
+    """Regression for #3301: a second-granularity stamp plus a 50-char slug is
+    not unique, so saves in the same second sharing a prefix collapsed into one
+    file and the earlier ones were silently lost."""
+    from concurrent.futures import ThreadPoolExecutor
+    mem = tmp_path / "memory"
+    with ThreadPoolExecutor(max_workers=20) as ex:
+        paths = list(ex.map(lambda _: save_query_result("how does auth work", "a", mem), range(20)))
+    assert len({p.name for p in paths}) == 20
+    assert len(list(mem.glob("*.md"))) == 20