Module search 

Fuzzy search over the graph — multi-signal ranking with CamelCase/snake_case tokenisation, token overlap scoring, and edit-distance typo tolerance.

Ranking signals (higher score = better match):

• exact name match: +100
• prefix name match: +60
• all query tokens match name tokens: +50
• substring in name: +30
• partial token overlap: +10..+25
• edit distance ≤1 (typo): +20
• edit distance ≤2: +10
• substring in qualified_name only: +10
• shorter names break ties
• kind boost: Function/Method/Struct/Trait > others

Structs

FileGroup

Hits grouped by file — lets a model see “parse is in parser.rs (8 symbols)” at a glance without iterating the full flat hit list.

SearchHit

Functions

group_by_file

Group search hits by file. Groups are sorted by symbol_count descending (most hits first), ties broken alphabetically by file path.

search

Run a fuzzy search across all nodes on branch.