bug: HTTP route heuristic misclassifies filesystem paths; client/server routes never join

[ilyabrykau-orca]

Describe the bug

On a Python repo (~82.6k graph nodes), route / HTTP_CALLS extraction produces false and unjoinable data:

• 1062 / 1213 Route nodes are empty URL-string-literal stubs (no method/route_path).
• url_path includes filesystem paths misclassified as HTTP — /root/.aws/credentials, /etc/crio/crio.conf — and str.split('/locations/') delimiters; os.remove/os.path.join are emitted as HTTP_CALLS.
• Client and server route sets are fully disjoint: Route<-[:HANDLES]- ∩ Route<-[:HTTP_CALLS]- = 0. No client call resolves to a handled endpoint; internal URLs (http://api-admin:8001) exist only as orphan Route name strings, 0 wired.

Expected

Route extraction distinguishes HTTP URLs from filesystem paths / string-split args; client HTTP_CALLS resolve to server Routes (intra-repo at minimum).

Repro

Index any repo mixing requests/HTTP clients with os.path/file I/O, then:

MATCH (a)-[:HTTP_CALLS]->(b:Route) RETURN b.url_path

Confirmations

• I searched existing issues. Distinct from Protocol-aware cross-repo intelligence: gRPC + typed-message + attribute-route matching (design proposal) #292 (cross-repo protocol matching) — this is intra-repo extraction correctness, and a prerequisite for Protocol-aware cross-repo intelligence: gRPC + typed-message + attribute-route matching (design proposal) #292's matcher to have clean Route nodes to work with.

[DeusData]

Thanks for the report — a route heuristic that misclassifies filesystem paths is a real graph-quality bug. Will investigate.