Fix #683: normalize source_file path separators + two-phase cohesion re-clustering

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

graphify/build.py

@@ -39,6 +39,12 @@ def _normalize_id(s: str) -> str:
     return cleaned.strip("_").lower()
 
 
+def _norm_source_file(p: str | None) -> str | None:
+    """Normalize path separators to forward slashes so Windows backslash paths
+    and POSIX paths from semantic subagents resolve to the same node identity."""
+    return p.replace("\\", "/") if p else p
+
+
 def build_from_json(extraction: dict, *, directed: bool = False) -> nx.Graph:
     """Build a NetworkX graph from an extraction dict.
 
@@ -73,6 +79,8 @@ def build_from_json(extraction: dict, *, directed: bool = False) -> nx.Graph:
         print(f"[graphify] Extraction warning ({len(real_errors)} issues): {real_errors[0]}", file=sys.stderr)
     G: nx.Graph = nx.DiGraph() if directed else nx.Graph()
     for node in extraction.get("nodes", []):
+        if "source_file" in node:
+            node["source_file"] = _norm_source_file(node["source_file"])
         G.add_node(node["id"], **{k: v for k, v in node.items() if k != "id"})
     node_set = set(G.nodes())
     # Normalized ID map: lets edges survive when the LLM generates IDs with
@@ -95,6 +103,8 @@ def build_from_json(extraction: dict, *, directed: bool = False) -> nx.Graph:
         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")}
+        if "source_file" in attrs:
+            attrs["source_file"] = _norm_source_file(attrs["source_file"])
         # Preserve original edge direction - undirected graphs lose it otherwise,
         # causing display functions to show edges backwards.
         attrs["_src"] = src

graphify/cluster.py

@@ -54,6 +54,8 @@ def _partition(G: nx.Graph) -> dict[str, int]:
 
 _MAX_COMMUNITY_FRACTION = 0.25   # communities larger than 25% of graph get split
 _MIN_SPLIT_SIZE = 10             # only split if community has at least this many nodes
+_COHESION_SPLIT_THRESHOLD = 0.05 # re-split communities with cohesion below this
+_COHESION_SPLIT_MIN_SIZE = 50    # only cohesion-split if community has at least this many nodes
 
 
 def cluster(G: nx.Graph) -> dict[int, list[str]]:
@@ -99,6 +101,17 @@ def cluster(G: nx.Graph) -> dict[int, list[str]]:
         else:
             final_communities.append(nodes)
 
+    # Second pass: re-split low-cohesion communities caused by doc-hub nodes
+    # that bridge otherwise-unrelated subsystems (e.g. CLAUDE.md connected to everything).
+    second_pass: list[list[str]] = []
+    for nodes in final_communities:
+        if len(nodes) >= _COHESION_SPLIT_MIN_SIZE and cohesion_score(G, nodes) < _COHESION_SPLIT_THRESHOLD:
+            splits = _split_community(G, nodes)
+            second_pass.extend(splits if len(splits) > 1 else [nodes])
+        else:
+            second_pass.append(nodes)
+    final_communities = second_pass
+
     # Re-index by size descending for deterministic ordering
     final_communities.sort(key=len, reverse=True)
     return {i: sorted(nodes) for i, nodes in enumerate(final_communities)}

tests/test_build.py

@@ -50,6 +50,21 @@ def test_legacy_edge_from_to_canonicalized():
     assert G.number_of_edges() == 1
 
 
+def test_source_file_backslash_normalized():
+    """Windows backslash paths and POSIX paths for the same file must produce one node."""
+    extraction = {
+        "nodes": [
+            {"id": "n1", "label": "A", "file_type": "code", "source_file": "src\\middleware\\auth.py"},
+            {"id": "n2", "label": "B", "file_type": "code", "source_file": "src/middleware/auth.py"},
+        ],
+        "edges": [],
+        "input_tokens": 0, "output_tokens": 0,
+    }
+    G = build_from_json(extraction)
+    sources = {G.nodes[n]["source_file"] for n in G.nodes()}
+    assert sources == {"src/middleware/auth.py"}
+
+
 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}