Distillation

🔴 (00:02) User stated “replacing markdown with truncated,” overriding the prior representation so section markdown is replaced with truncated content.
🔴 (00:02) User-provided changelog code imports marked, Token, and Tokens from "marked" and defines VERSION_PREFIX_RE = /^v/.
🔴 (00:02) User-provided changelog code defines ChangeCategory, ChangeSection, and a HEADING_TO_CATEGORY: ReadonlyMap<string, ChangeCategory> for mapping normalized headings to categories.
🔴 (00:02) User-provided heading normalization strips emoji and symbols, removes extra whitespace, and lowercases heading text for category matching.
🔴 (00:02) User-provided extractSections(body: string): ChangeSection[] parses changelog content into categorized sections.
🔴 (00:02) User-provided list-processing logic counts top-level items using list.items.length and slices according to that top-level count.
🔴 (00:02) User-provided truncation logic mutates the sections array in place and returns { totalItems, truncated: totalItems < originalCount }; its item budget is clamped with Math.max(0, budget).
🔴 (00:02) User-provided changelog aggregation stores sections by category using sectionsByCategory.get(section.category) ?? [].
🔴 (00:02) User-provided release filtering computes inRange releases between version bounds, with toVersion documented as the inclusive upper bound.
🔴 (00:02) User-provided commit parsing uses a conventional-commit prefix-to-category mapping; only feat:, fix:, and perf: prefixes pass through, and the description is derived from message.split("\n")[0]?.trim().
🔴 (00:02) User-provided release fetching returns releases newest first and returns an empty array on failure.
🔴 (00:02) User-provided changelog fetching supports maxItems?: number; GitHub’s inclusive since boundary is offset by +1s to exclude the current version’s commit, and changelog generation can return null on fetch failure or invalid versions.
🔴 (00:02) User-provided top-level async entry point is fetchChangelog.

Distillation

Observations