Struct SearchConfig 

pub struct SearchConfig {
    pub batch_size: usize,
    pub max_tokens: usize,
    pub chunk: ChunkConfig,
    pub text_mode: bool,
    pub cascade_dim: Option<usize>,
    pub file_type: Option<String>,
    pub exclude_extensions: Vec<String>,
    pub include_extensions: Vec<String>,
    pub ignore_patterns: Vec<String>,
    pub scope: Scope,
    pub mode: SearchMode,
}

Runtime configuration for the search pipeline.

All tuning parameters that were previously compile-time constants are gathered here so they can be set from CLI arguments without recompiling.

Fields

batch_size: usize

Chunks per inference call. Larger values amortize call overhead but consume more memory. Default: 32.

max_tokens: usize

Maximum tokens fed to the model per chunk. 0 means no limit. Capping tokens controls inference cost for minified or dense source. BERT attention cost scales linearly with token count, and CLS pooling means the first token’s representation carries most semantic weight. Default: 128 (7.7× faster than 512, with minimal quality loss).

chunk: ChunkConfig

Chunking parameters forwarded to the chunking phase.

text_mode: bool

Force all files to be chunked as plain text (sliding windows only). When false (default), files with recognized extensions use tree-sitter semantic chunking, and unrecognized extensions fall back to sliding windows.

cascade_dim: Option<usize>

MRL cascade pre-filter dimension.

When set, SearchIndex stores a truncated and L2-re-normalized copy of the embedding matrix at this dimension for fast two-phase cascade search. None (default) disables cascade search.

file_type: Option<String>

Optional file type filter (e.g. “rust”, “python”, “js”).

When set, only files matching this type (using ripgrep’s built-in type database) are collected during the walk phase.

exclude_extensions: Vec<String>

File extensions to exclude during the walk phase.

include_extensions: Vec<String>

File extensions to include during the walk phase. Empty means “no extension whitelist” (other filters still apply). Non-empty restricts walking to files whose extension matches one of these (normalized lowercase, with or without leading dot).

ignore_patterns: Vec<String>

Additional .gitignore-style patterns to exclude during the walk phase.

scope: Scope

Intent-shaped scope: code, docs, or all. Drives the default extension whitelist when include_extensions is empty and the rerank gate in the MCP layer (docs and all-on-mixed-corpus fire rerank; code skips). See Scope.

mode: SearchMode

Search mode: hybrid (default), semantic, or keyword.

Implementations

impl SearchConfig

pub fn walk_options(&self) -> WalkOptions

Convert search configuration into shared walk filters.

Resolves the scope-implied extension whitelist:

• Explicit include_extensions always wins.
• Otherwise Scope::Docs injects the canonical prose set (PROSE_EXTENSIONS).
• Scope::Code injects the canonical prose set as exclusions (so prose files are skipped during walk).
• Scope::All leaves the include set empty (no whitelist).

pub fn apply_repo_config(&mut self, root: &Path)

Merge ignore patterns from .ripvec/config.toml, if present.

Trait Implementations

impl Clone for SearchConfig

fn clone(&self) -> SearchConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for SearchConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for SearchConfig

fn default() -> Self

Returns the “default value” for a type.