Distillation

Date: Sep 10, 2026

• 🔴 (03:05) User-provided GHCR source imports getUserAgent from ./constants.js and supports the standard ghcr.io anonymous token exchange.
• 🔴 (03:05) User-provided GHCR source notes that AbortSignal.timeout() throws a TimeoutError DOMException and checks the error by name.
• 🔴 (03:05) User-provided GHCR request logic uses options?.timeout ?? GHCR_REQUEST_TIMEOUT, supports an external AbortSignal, and stores failures as error instanceof Error ? error : new Error(String(error)).
• 🔴 (03:05) User-provided GHCR source defines GHCR_REGISTRY as the base URL for the GHCR registry API.
• 🔴 (03:05) User-provided GHCR source represents every binary in a nightly push as a separate OCI layer and supports manifest annotations?: Record<string, string> plus a numeric schemaVersion.
• 🔴 (03:05) User-provided GHCR source fetches a short-lived anonymous bearer token for read-only registry access and throws UpgradeError for network failures or non-200 responses.
• 🔴 (03:05) User-provided GHCR source includes findLayerByFilename() to locate a binary layer.
• 🔴 (03:05) User-provided GHCR blob download constructs ${GHCR_REGISTRY}/v2/${source.ghcrRepo}/blobs/${digest}, initially uses redirect: "manual", and accepts either a direct HTTP 200 response or redirects with status 301, 302, 307, or 308.
• 🔴 (03:05) User-provided GHCR source rejects a redirect response without a redirect URL and explains that an unsuitable timeout would require sustained approximately 8 Mbps throughput and fail on slow connections.
• 🔴 (03:05) User-provided GHCR tag listing can throw UpgradeError detail Failed to list GHCR tags: HTTP ${response.status}; paginated tags are accumulated in allTags and returned in the registry’s lexicographic order.
• 🔴 (03:05) User-provided changelog source imports valid as semverValid from "semver" and defines VERSION_PREFIX_RE = /^v/.
• 🔴 (03:05) User-provided changelog parsing strips emoji and symbols from heading text for category matching, normalizes headings through normalizeHeading(), and maps "new features" to "features".
• 🔴 (03:05) User-provided changelog section data preserves raw Markdown and tracks the total number of list items across all sections.
• 🔴 (03:05) User-provided changelog logic mutates the sections array in place, replacing markdown with truncated content.
• 🔴 (03:05) User-provided changelog truncation uses a maxItems: number limit and decrements a remaining item budget by list.items.length.
• 🔴 (03:05) User-provided changelog logic groups content with sectionsByCategory.get(category) and normalizes release data through normalizeChangelogReleases().
• 🔴 (03:05) User-provided changelog fetching uses a higher per_page value than fetchRecentReleases() in the delta-upgrade implementation.
• 🔴 (03:05) User-provided changelog logic can normalize prefetchedReleases via normalizeChangelogReleases(prefetchedReleases, source, true) and includes nightly changelog handling based on commits: GitHubCommit[] and a toVersion target nightly version.