Crate apohara_indexer 

Apohara code indexer (soft-fork): sqlite-vec storage + blake3 feature-hashing embeddings + tree-sitter parsing.

This crate is a LIB-ONLY soft-fork of apohara-indexer from SuarezPM/Apohara-Catalyst. The storage, parser, and embeddings modules are preserved verbatim; the binary entry point (main.rs) is intentionally dropped.

The new engine modules land in subsequent steps. Step 1 adds walker (gitignore-aware traversal) and chunker (symbol/module/window chunking); schema/search/incremental follow later.

Re-exports

pub use storage::ensure_vec_extension_registered;

pub use storage::insert_chunk_full;

pub use storage::insert_chunk_full_with;

pub use storage::knn_query;

pub use storage::knn_query_with;

pub use storage::open_db;

pub use storage::open_db_with;

pub use storage::write_file_structural;

pub use storage::IndexedChunk;

pub use storage::KnnHit;

pub use storage::SymbolData;

pub use storage::EMBED_DIM;

pub use embedder::active_embedder;

pub use embedder::resolve_embedder_choice;

pub use embedder::Embedder;

pub use embedder::EmbedderChoice;

pub use embedder::FeatureHashEmbedder;

pub use embedder::EMBED_MODEL_ENV;

pub use embedder::FEATURE_HASH_ID;

pub use registry::load as load_registry;

pub use registry::register;

pub use registry::registry_path;

pub use registry::save as save_registry;

pub use registry::Registry;

pub use schema::migrate;

pub use schema::read_embedder_meta;

pub use schema::verify_embedder_meta;

pub use schema::write_embedder_meta;

pub use schema::META_EMBEDDER_DIM;

pub use schema::META_EMBEDDER_ID;

pub use schema::META_SCHEMA_VERSION;

pub use schema::MIGRATION_PLACEHOLDER_REPO_ID;

pub use schema::SCHEMA_VERSION;

pub use search::apply_structural_boost;

pub use search::bm25_query;

pub use search::classify_query_weights;

pub use search::dedup_content;

pub use search::dedup_overlapping;

pub use search::hydrate;

pub use search::load_embeddings;

pub use search::mmr_rerank;

pub use search::resolve_weights;

pub use search::rrf_fuse;

pub use search::rrf_fuse_weighted;

pub use search::vector_query;

pub use search::vector_query_with;

pub use search::ExportRow;

pub use search::HydratedHit;

pub use search::ImportRow;

pub use search::MMR_LAMBDA;

pub use search::RRF_K;

pub use search::STRUCTURAL_BOOST;

pub use tokens::code_tokens;

pub use embeddings::feature_hash_embed;

pub use parser::detect_language;

pub use parser::parse_file;

pub use parser::parse_imports_exports;

pub use parser::parse_source;

pub use parser::parse_source_imports_exports;

pub use parser::parse_source_spans;

pub use parser::ExportStatement;

pub use parser::FunctionSignature;

pub use parser::ImportStatement;

pub use parser::Language;

pub use parser::SymbolKind;

pub use walker::walk_repo;

pub use walker::WalkedFile;

pub use chunker::chunk_file;

pub use chunker::chunk_id;

pub use chunker::ChunkKind;

pub use chunker::ChunkSpec;

pub use incremental::index_repo;

pub use incremental::index_repo_with;

pub use incremental::reindex;

pub use incremental::reindex_with;

pub use incremental::ReindexReport;

Modules

chunker

Splits file contents into indexable chunks.

embedder

Pluggable embedding behind the Embedder trait.

embeddings

Deterministic feature-hashing embeddings (blake3-based).

incremental

Incremental (and full) reindex engine.

parser

registry

Multi-repo sidecar registry: a path -> index.db map (Decision E1).

schema

Idempotent schema migration run on every open AFTER crate::storage::open_db.

search

Query-time retrieval: lexical (BM25 over FTS5), vector (sqlite-vec KNN), Reciprocal Rank Fusion of the two, and hydration of a fused hit into a displayable record.

storage

sqlite-vec backed storage for code chunks + their embeddings.

tokens

Code-aware tokenizer shared by the index-time and query-time paths.

walker

Gitignore-aware filesystem traversal for indexing.

Functions

sqlite_version

Bundled SQLite version string (e.g. "3.46.0").