Enum CorpusClass 

pub enum CorpusClass {
    Code,
    Mixed,
    Docs,
}

Index-time classification of the corpus by file mix.

Drives the corpus-aware rerank gate: docs and mixed corpora get the L-12 cross-encoder fired (when the query is NL-shaped); pure code corpora skip it because the ms-marco-trained model is out-of-domain for code regardless of impl quality.

Variants

Code

Less than 30% of chunks are in prose files. Pure or near-pure code corpora — rerank skipped.

Mixed

Between 30% and 70% prose chunks. Mixed corpora — rerank fires on NL queries to recover the prose-dominant relevance signal.

Docs

At least 70% prose chunks. Documentation, book sets, knowledge bases — rerank fires by default.

Implementations

impl CorpusClass

pub fn classify(chunks: &[CodeChunk]) -> Self

Classify a chunk set by the fraction of chunks dominated by prose.

A chunk counts as “prose” when either of the following holds:

• its file extension is in crate::encoder::ripvec::ranking::is_prose_path (e.g. .md, .rst, .txt), OR
• its content is dominated by docstring/comment text per [chunk_is_prose_dominated] (the I#64 / B-0028 path — docstring-heavy Python, JS-doc-heavy code, etc.).

The second branch matters: a Mnemosyne-class Python corpus where every class has a substantial docstring is classified as Code by the file-extension test alone, even though >50% of its bytes are prose. The within-chunk content test catches this case so Auto-policy rerank fires on NL queries against such corpora.

Empty input is classified as Code (degenerate but defined).

pub fn rerank_eligible(self) -> bool

Whether the cross-encoder rerank should run on this corpus for a non-symbol NL query. Pure code corpora skip rerank; mixed and docs corpora enable it.

Trait Implementations

impl Clone for CorpusClass

fn clone(&self) -> CorpusClass

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for CorpusClass

impl Debug for CorpusClass

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for CorpusClass

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for CorpusClass

impl PartialEq for CorpusClass

fn eq(&self, other: &CorpusClass) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for CorpusClass

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for CorpusClass