internal/

Package chunk defines the common chunk shape produced by every chunker (markdown and tree-sitter) so the downstream embedding/storage path is uniform regardless of source language.

Package config loads the repository's .pkb/config.toml, which selects an embedding model and optional extension routing overrides.

Package cost holds the pricing table and unit conversions used to estimate the dollar cost of a reindex run before any paid embedding work is done.

Package embed defines the embedding-model interface used to embed all files (code and text) via a single model.

Package filetype routes a file extension to a FileType (code or text) and, for code, the tree-sitter grammar name used to chunk it.

Package git wraps the subset of git plumbing PKB needs to discover files and detect changes: repo root resolution, ref/sha resolution, tree listing, and materializing the staging area as a tree.

Package index implements the reindex flow: it lists the target tree (HEAD by default) and compares each file's git blob sha against the blob shas already recorded in the store, then indexes/updates/deletes only the files that differ.

Package mirror defines the on-disk mirror-artifact format: the committed, per-source-file index record split across two sibling files.

Package paths defines nominal path types so the compiler distinguishes a path relative to the git repository root (as produced by git plumbing and stored in the index) from an absolute filesystem path (used when touching the working tree or reporting results to a user).

Package querylog appends every search query to a JSONL file under .pkb, so the accumulated log can later seed a search eval/validation suite.

Package store wraps the sqlite + sqlite-vec database: a files table keyed by (path, model, version), a chunks table, and one vec0 table per (modelName, version).