Feature request: scoped resolution for explain/query/path (monorepos, minified bundles, name collisions)

[DevangML]

Environment

• graphifyy: PyPI graphifyy (CLI graphify); version pinned at time of report e.g. 0.5.x
• Python: 3.10+
• Corpus: Large multi-root workspace: multiple Frappe benches (same app names under different roots), frappe / erpnext trees, and generated assets under sites/assets/ (including *.min.js).
• Command: graphify update . from repository root

Problem

For this corpus shape, the defaults for explain, graphify query, and graphify path are easy to misread:

1. explain "<label>" often attaches to the wrong node for short or overloaded names — e.g. symbols inside minified sites/assets/**/*.min.js, or a test class before the Document controller for the same name.
2. query (BFS) surfaces high-degree framework hubs (get, get_doc, Document, ERPNext-heavy nodes) before app-specific nodes unless --budget is very large.
3. path "A" "B" often returns no path for informal strings; disambiguation is needed when the same relative path exists under multiple bench roots.

graph.json itself still contains the right per-file nodes and many EXTRACTED edges; the pain is mostly resolution / traversal ergonomics and signal-to-noise at the CLI, not “missing data” in the graph.

Workarounds (already supported — README)

These reduce noise before any new flags:

• .graphifyignore (gitignore-style) — e.g. **/sites/assets/, **/*.min.js; optionally exclude frappe/ / erpnext/ when analysis targets custom apps only.
• Narrow scan root — graphify update <subdir> (e.g. one frappe-bench/apps/ tree).
• merge-graphs — combine smaller graphs when a single full-root graph is too hub-heavy.

This issue is for cases where teams still want a wide index or need better disambiguation without hand-editing graph.json.

Proposal

Idea	Rationale
explain --file <path> [--line N] and/or top-k matches (each with source_file, line, node id)	Avoids a single silent wrong match; matches how power users already use stable ids from graph.json.
query --seed-file <path> or --path-prefix <prefix>	Starts BFS from nodes under a subtree (one bench / one app), reducing hub flooding.
Optional deprioritization of certain paths as match / seed candidates (not only via ignore)	For corpora that must include sites/assets but should not prefer minified bundles when resolving validate / get_doc.

Suggested acceptance (future behavior)

• With explain --file (or equivalent) set to a known controller, resolution for a short name like validate should not prefer an unrelated *.min.js symbol in that disambiguated context.
• explain DataSyncConfiguration-style input should return multiple candidates (e.g. test vs Document) with full source_file, not pick one arbitrarily.

Related

• Cross-file INFERRED calls resolution lacks import evidence + ambiguous-candidate handling — short common names blow up god_nodes ranking #543 — INFERRED cross-file calls / global name map; short names and god_nodes inflation (overlapping theme: ambiguous short names).
• README — .graphifyignore, merge-graphs, query / path / explain examples.

Happy to help validate a PR or share a minimal public repro corpus description if useful.