Conventional Release Notes Builder

How to use Conventional Release Notes Builder

Turn supplied commit messages into a reviewable grouped Markdown release draft with exact source-record evidence. Collect the feature, fix and breaking markers already present in your own messages, then inspect what entered or missed the draft. The browser does not open a repository, infer a release range, select a version or publish a changelog; the separate artifact remains yours to review.

1. Provide optional authored commit json, release title, date and commit collection json.
2. Choose the supported mode and run the tool. Review its assumptions and any reported errors.
3. Copy the result or download a separate output; reset to clear the current form.

Good to know

• One strict UTF-8 JSON model up to 100,000 bytes, 10,000 values and 40 levels. Root keys are title, date, commits. Duplicate/unsafe members and unknown keys refuse; a selected file takes precedence over paste. Title is a bounded 160-byte single-line string, and date is empty or an actual ISO Gregorian YYYY-MM-DD value in years 0001–9999. Neither establishes a real release.
• Supply 1–200 objects with message and an optional id. Each message is at most 10,000 UTF-8 bytes. IDs, when provided, are exact unique 7–64 hexadecimal tokens under ASCII case folding; they are not checked against Git or turned into invented repository links.
• The existing selected Conventional Commit parser separates a bounded ASCII type, optional scope/bang, description, blank-line body and footer candidates. Uppercase BREAKING CHANGE/BREAKING-CHANGE receive their reserved meaning. CRLF becomes LF; NUL, bare CR, ambiguous or unsupported syntax can refuse. Review each retained parser warning and exact body/footer evidence.
• Refuse invalid is atomic: one unparsed message prevents any draft. Keep diagnostics excludes unparsed messages and records their source indexes/reasons. Invalid overall model/encoding/IDs/size still refuse the whole input. No absent commits, diffs or repository facts are fabricated.
• Each parsed message appears exactly once: breaking, then nonbreaking feat, nonbreaking fix, other. Original order remains inside each group. Other types stay visible; breaking footer line breaks are collapsed to spaces only in Markdown details, while exact parsed evidence remains in JSON.
• All title/scope/description/footer text is escaped as inert Markdown source. The tool page never renders or executes the draft, HTML or commit bodies. The Markdown artifact is at most 200,000 UTF-8 bytes. Reports can contain private commit text; no automatic save, upload, Git command, hook or network access occurs.
• The descriptive major/minor/patch/none message signal is a convention summary, not a semantic-compatibility judgment, version recommendation, 0.x policy, release date, deployment or publication. Check actual code changes and your release process before using the generated draft.
• Structured text output is limited to 500,000 characters. Reduce the supplied source when that separate output budget is exceeded.
• Inputs are processed on this device and are not automatically saved or uploaded.