fix: classify repeated edge occurrences

CHANGELOG.md

@@ -2,6 +2,26 @@
 
 Full release notes with details on each version: [GitHub Releases](https://github.com/safishamsi/graphify/releases)
 
+## Unreleased
+
+- Fix: AST edges now retain portable token-level occurrence evidence. Python
+  annotations use their real line/column span and parameter name, repeated
+  evidence on one relation is aggregated deterministically, and duplicate
+  emissions introduced by post-resolution endpoint rewrites are suppressed.
+  Diagnostics separate legitimate source repetition from producer defects.
+  Deliberately unindexed data JSON is now a cached intentional skip rather than
+  a recurring zero-node warning, with failures and unsupported files reported
+  separately.
+- Feat: add opt-in `--multigraph` builds backed by a persistent directed
+  `MultiDiGraph`. Canonically distinct relations between the same endpoints
+  retain stable keys; exact duplicate occurrences aggregate through
+  `occurrence_count`. Incremental updates, watch/hook rebuilds, cluster-only,
+  no-cluster, graph merges, query/path/explain, HTML, wiki, Obsidian, GraphML,
+  graph databases, MCP, reports, and diagnostics preserve or explicitly reject
+  unsupported simplification. Community detection, cohesion, god nodes, and
+  centrality use an unweighted simple undirected topology projection so
+  multiplicity does not distort analysis. The default remains `Graph`.
+
 ## 0.9.26 (2026-07-25)
 
 - Fix: `graphify query`/`explain` no longer fabricate `indirect_call` edges to class definitions (#2137, thanks @Rishet11). The callable guard admitted classes, so passing a class as a value (`select(Model)`, `db.get(Model, id)`, `except (ErrorA, ErrorB)`, `getattr(obj, "Name", 0)`) produced a false inferred call edge in both the intra-file and cross-file paths. Classes are now tracked separately and excluded from `indirect_call`; direct instantiation still emits its `calls` edge. The suppression is context-blind, so a genuine higher-order class callback that is actually invoked (e.g. `map(Point, coords)`) also loses its edge, which is the intended tradeoff.

README.md

@@ -734,6 +734,7 @@ graphify extract ./docs --api-timeout 900      # longer HTTP timeout for slow lo
 graphify extract ./docs --google-workspace     # export .gdoc/.gsheet/.gslides via gws before extraction
 graphify extract ./src --no-gitignore          # include git-ignored source; still honor .graphifyignore
 graphify extract ./docs --mode deep            # richer semantic extraction via extended system prompt
+graphify extract ./src --multigraph            # preserve directed parallel relations with stable edge keys
 graphify extract ./docs --no-cluster           # raw extraction only, skip clustering
 graphify extract ./docs --timing               # print per-stage wall-clock timings to stderr (also works on cluster-only)
 graphify extract ./docs --force                # overwrite graph.json even if new graph has fewer nodes (use after refactors or to clear ghost duplicates)

graphify/__main__.py

@@ -521,13 +521,15 @@ def _run_cli() -> None:
         print("    --max-examples N        max same-endpoint examples to print (default 5)")
         print("    --directed              force directed post-build simulation")
         print("    --undirected            force undirected post-build simulation")
+        print("    --multigraph            force MultiDiGraph post-build simulation")
         print("                            (default follows JSON directed flag;")
         print("                             raw extraction with no flag defaults directed)")
         print("    --extract-path PATH     extractor source for suppression scan")
         print("  clone <github-url>      clone a GitHub repo locally and print its path for /graphify")
         print("  merge-driver <base> <current> <other>  git merge driver: union-merge two graph.json files (set up via hook install)")
         print("  merge-graphs <g1> <g2>  merge two or more graph.json files into one cross-repo graph")
         print("    --out <path>            output path (default: graphify-out/merged-graph.json)")
+        print("    --multigraph            preserve parallel relations in a directed multigraph")
         print("    --branch <branch>       checkout a specific branch (default: repo default)")
         print("    --out <dir>             clone to a custom directory (default: ~/.graphify/repos/<owner>/<repo>)")
         print("  add <url>               fetch a URL and save it to ./raw, then update the graph")
@@ -536,6 +538,7 @@ def _run_cli() -> None:
         print("    --dir <path>            target directory (default: ./raw)")
         print("  watch <path>            watch a folder and rebuild the graph on code changes")
         print("  update <path>           re-extract code files and update the graph (no LLM needed)")
+        print("    --multigraph            preserve parallel directed relations (also preserves existing mode)")
         print("    --force                 overwrite graph.json even if the rebuild has fewer nodes")
         print("                            (also: GRAPHIFY_FORCE=1 env var; use after refactors that delete code)")
         print("    --no-cluster            skip clustering, write raw extraction only")
@@ -609,6 +612,7 @@ def _run_cli() -> None:
         print("    --api-timeout S         per-request timeout in seconds for the LLM client (default: 600)")
         print("    --out DIR, --output DIR output dir (default: <path>); writes <DIR>/graphify-out/")
         print("    --google-workspace      export .gdoc/.gsheet/.gslides shortcuts via gws before extraction")
+        print("    --multigraph            preserve parallel directed relations in a MultiDiGraph")
         print("    --no-gitignore         ignore .gitignore and .git/info/exclude (prioritizes .graphifyignore)")
         print("    --no-cluster            skip clustering, write raw extraction only")
         print("    --code-only             index code (local AST, no API key) and skip doc/paper/image files")

graphify/analyze.py

@@ -3,7 +3,7 @@
 from pathlib import Path
 import networkx as nx
 
-from graphify.build import edge_data
+from graphify.build import analysis_projection, edge_data
 
 # Builtin/mock names that can appear as annotation-derived nodes in pre-existing
 # graphs. Excluded from god-node ranking so they don't displace real abstractions
@@ -104,7 +104,8 @@ def god_nodes(G: nx.Graph, top_n: int = 10) -> list[dict]:
     File-level hub nodes are excluded: they accumulate import/contains edges
     mechanically and don't represent meaningful architectural abstractions.
     """
-    degree = dict(G.degree())
+    topology = analysis_projection(G)
+    degree = {node: topology.degree(node) for node in G.nodes}
     sorted_nodes = sorted(degree.items(), key=lambda x: x[1], reverse=True)
     result = []
     for node_id, deg in sorted_nodes:
@@ -281,7 +282,7 @@ def _cross_file_surprises(G: nx.Graph, communities: dict[int, list[str]], top_n:
     Each result includes a 'why' field explaining what makes it non-obvious.
     """
     node_community = _node_community_map(communities)
-    degrees = dict(G.degree())
+    degrees = dict(analysis_projection(G).degree())
     candidates = []
 
     for u, v, data in G.edges(data=True):
@@ -347,7 +348,8 @@ def _cross_community_surprises(
             return []
         if G.number_of_nodes() > 5000:
             return []
-        betweenness = nx.edge_betweenness_centrality(G)
+        topology = analysis_projection(G)
+        betweenness = nx.edge_betweenness_centrality(topology)
         top_edges = sorted(betweenness.items(), key=lambda x: x[1], reverse=True)[:top_n]
         result = []
         for (u, v), score in top_edges: