Backfill edge source_file from endpoint nodes (#1279)

Semantic/LLM edges occasionally omit source_file, which build only normalized
when already present, so the field reached graph.json empty and downstream
validation flagged it. Backfill from the source (then target) node in
build_from_json and in the --no-cluster raw-write path, which bypasses it.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

graphify/__main__.py

@@ -4453,6 +4453,14 @@ def main() -> None:
 
             merged["nodes"] = _dedupe_nodes(merged["nodes"])
             merged["edges"] = _dedupe_edges(merged["edges"])
+            # Backfill source_file from endpoint nodes — this raw path bypasses
+            # build_from_json's backfill, and semantic edges sometimes omit it (#1279).
+            _node_sf = {n.get("id"): n.get("source_file") for n in merged["nodes"]}
+            for _e in merged["edges"]:
+                if not _e.get("source_file"):
+                    _e["source_file"] = (
+                        _node_sf.get(_e.get("source")) or _node_sf.get(_e.get("target")) or ""
+                    )
             _backup(graphify_out)
             graph_json_path.write_text(
                 json.dumps(merged, indent=2), encoding="utf-8"

graphify/build.py

@@ -296,6 +296,15 @@ def build_from_json(extraction: dict, *, directed: bool = False, root: str | Pat
         if src not in node_set or tgt not in node_set:
             continue  # skip edges to external/stdlib nodes - expected, not an error
         attrs = {k: v for k, v in edge.items() if k not in ("source", "target")}
+        # Backfill source_file from the endpoint nodes (every node carries one).
+        # Semantic/LLM edges occasionally omit it, which downstream validation
+        # flags and leaves query results with no file reference (#1279).
+        if not attrs.get("source_file"):
+            attrs["source_file"] = (
+                G.nodes[src].get("source_file")
+                or G.nodes[tgt].get("source_file")
+                or ""
+            )
         if "source_file" in attrs:
             attrs["source_file"] = _norm_source_file(attrs["source_file"], _root)
         # Drop cross-language INFERRED `calls` edges — same short names (render,

tests/test_build.py

@@ -108,6 +108,23 @@ def test_source_file_backslash_normalized():
     assert sources == {"src/middleware/auth.py"}
 
 
+def test_edge_missing_source_file_backfilled_from_node():
+    """#1279: a semantic/LLM edge lacking source_file must inherit it from its
+    source node rather than reach graph.json with no file reference."""
+    extraction = {
+        "nodes": [
+            {"id": "n1", "label": "A", "file_type": "concept", "source_file": "docs/a.md"},
+            {"id": "n2", "label": "B", "file_type": "concept", "source_file": "docs/b.md"},
+        ],
+        # No source_file on the edge (as LLM output sometimes omits it).
+        "edges": [{"source": "n1", "target": "n2", "relation": "relates_to", "confidence": "INFERRED"}],
+        "input_tokens": 0, "output_tokens": 0,
+    }
+    G = build_from_json(extraction)
+    sf = edge_data(G, "n1", "n2").get("source_file")
+    assert sf == "docs/a.md"  # backfilled from the source node
+
+
 def test_build_merges_multiple_extractions():
     ext1 = {"nodes": [{"id": "n1", "label": "A", "file_type": "code", "source_file": "a.py"}],
             "edges": [], "input_tokens": 0, "output_tokens": 0}