internal/

Package accuracy is the corpus scoring harness: it runs the full Phase 11 pipeline against a set of recorded filings (hermetic — the corpus ships the bytes, nothing touches the network) and scores the assembled documents against hand-verified baselines.

Package cache is a two-layer, pure-Go SQLite store that makes repeated extraction cheap and keeps sec-cli a good EDGAR citizen by not re-fetching bytes it already has.

Package diff compares two normalized documents and emits a structured change set: numeric statement deltas matched by concept (labels drift, concepts are stable) and narrative changes matched by item id and diffed at paragraph granularity.

Package edgar is the HTTP client layer for SEC EDGAR.

Package ixbrl parses inline-XBRL filings: the fact stream and its contexts (this file's siblings), and — in later phases — the presentation-linkbase projection, layout tables, and free-text sections.

Package model is the normalized, source-agnostic output contract for a filing: the Document and the value types every other package produces, renders, caches, diffs, or scores.

Package pipeline is the conductor: the one place that knows the full ticker → fetch → detect → parse → project → partition → assemble sequence and returns a normalized *model.Document.

Package render turns a model.Document into one of three serializations behind a single Renderer interface: canonical JSON (the schema-stable contract the cache and Python wrapper read), Markdown (the LLM- and human-facing view), and plain text.

Package router classifies a filing's raw bytes into one of the EDGAR format eras.

Package sections partitions a filing into its named 10-K item sections and renders clean free text for each, with all inline-XBRL scaffolding removed.