Struct SymbolIndex 

pub struct SymbolIndex { /* private fields */ }

SQLite-backed symbol index for a project.

Architecture: writer Mutex<IndexDb> for mutations + per-query read-only connections for _cached methods. This makes SymbolIndex: Send + Sync, enabling Arc<SymbolIndex> without an external Mutex.

Implementations

impl SymbolIndex

pub fn find_symbol_cached( &self, name: &str, file_path: Option<&str>, include_body: bool, exact_match: bool, max_matches: usize, ) -> Result<Vec<SymbolInfo>>

Query symbols from DB without lazy indexing. Returns empty if file not yet indexed.

pub fn get_symbols_overview_cached( &self, path: &str, _depth: usize, ) -> Result<Vec<SymbolInfo>>

Get symbols overview from DB without lazy indexing.

pub fn get_ranked_context_cached( &self, query: &str, path: Option<&str>, max_tokens: usize, include_body: bool, depth: usize, graph_cache: Option<&GraphCache>, semantic_scores: HashMap<String, f64>, ) -> Result<RankedContextResult>

Ranked context from DB without lazy indexing. If graph_cache is provided, PageRank scores boost symbols in highly-imported files. If semantic_scores is non-empty, vector similarity is blended into ranking.

pub fn get_ranked_context_cached_with_query_type( &self, query: &str, path: Option<&str>, max_tokens: usize, include_body: bool, depth: usize, graph_cache: Option<&GraphCache>, semantic_scores: HashMap<String, f64>, query_type: Option<&str>, ) -> Result<RankedContextResult>

Like get_ranked_context_cached but accepts an optional query type ("identifier", "natural_language", "short_phrase") to tune ranking weights per query category.

pub fn get_ranked_context_cached_with_lsp_boost( &self, query: &str, path: Option<&str>, max_tokens: usize, include_body: bool, depth: usize, graph_cache: Option<&GraphCache>, semantic_scores: HashMap<String, f64>, query_type: Option<&str>, lsp_boost_refs: HashMap<String, Vec<usize>>, lsp_signal_weight: Option<f64>, ) -> Result<RankedContextResult>

Full form that additionally accepts per-file LSP reference line information for the P1-4 per-symbol boost.

lsp_boost_refs maps project-relative file paths to the list of reference lines that a caller-side reference probe (LSP textDocument/references, unioned with the tree-sitter text search) returned for the query’s target symbol. Inside rank_symbols, each candidate is boosted by how close its declaration line is to the nearest ref line at or below it — file-level uniform boost used to over-promote unrelated helpers in large ref’d files, per-symbol proximity keeps the lift targeted on plausible containers.

lsp_signal_weight scales the final boost. An empty ref map OR a None weight keeps the blend byte-identical to the pre-P1-4 pipeline.

impl SymbolIndex

pub fn refresh_all(&self) -> Result<IndexStats>

pub fn index_files(&self, paths: &[PathBuf]) -> Result<usize>

Incrementally re-index only the given files (changed/created).

pub fn refresh_file(&self, relative_path: &str) -> Result<usize>

Re-index a single file by relative path (convenience for post-mutation refresh).

pub fn remove_files(&self, paths: &[PathBuf]) -> Result<usize>

Remove deleted files from the index.

impl SymbolIndex

pub fn new(project: ProjectRoot) -> Self

pub fn new_memory(project: ProjectRoot) -> Self

Create an in-memory index (for tests and benchmarks — no disk persistence).

pub fn stats(&self) -> Result<IndexStats>

pub fn get_project_structure(&self) -> Result<Vec<DirStats>>

Hierarchical project structure: per-directory file count + symbol count. Used as Level 1 pruning — lets LLM decide which directories to drill into.

pub fn indexed_file_paths(&self) -> Result<Vec<String>>

pub fn get_symbols_overview( &self, path: &str, depth: usize, ) -> Result<Vec<SymbolInfo>>

pub fn find_symbol( &self, name: &str, file_path: Option<&str>, include_body: bool, exact_match: bool, max_matches: usize, ) -> Result<Vec<SymbolInfo>>

pub fn get_ranked_context( &self, query: &str, path: Option<&str>, max_tokens: usize, include_body: bool, depth: usize, ) -> Result<RankedContextResult>

pub fn db(&self) -> MutexGuard<'_, IndexDb>

Access the underlying database (e.g. for import graph queries).