leech

Templated everything

Filenames, file contents, and commit messages are all Nunjucks templates. One submission maps to one commit and can produce any number of files, in any directory layout.

Markdown & Typst

The raw problem HTML is exposed as question.content. Convert it with toMarkdown (GFM tables, strikethrough, task lists) or toTypst (real super/subscripts, escaping, code fences).

Self-contained repositories

Images in problem descriptions are downloaded into the repo and their references rewritten, so your solutions render offline and don't rot when a CDN URL changes.

Sync only what you want

Filter by status, language, problem, difficulty, tags, and a time window. Keep the noise out of your history.

Hooks at every step

Run your own shell commands before a sync, per submission (and mutate exactly what gets committed), or once at the end to regenerate an index.

Atomic, one commit per submission

Commits are built through the GitHub API without a checkout, then pushed in a single ref update. A run either lands every submission or none.

What it is

leech logs into LeetCode with your session cookies and pulls your accepted submissions through LeetCode's GraphQL API. For each one it renders a set of files from a template you control, commits them with the submission's own timestamp, and pushes once. There is no state file, no database, and no checkout of the target repository.

It's designed for people who want their solved problems to show up in a repo — as a GitHub profile flex, a study log, a blog source tree, or a Typst/PDF workbook — without a bespoke sync script.

Try it safely first

Every run supports dry-run: "true", which renders and logs everything but creates no commits. Build the whole pipeline locally with the CLI before you wire up the workflow.

How it works in one picture

LeetCode GraphQL                     GitHub API (no checkout)
────────────────                     ────────────────────────
submissionList  ──► filters ──┐
                              │
     for each submission:     │
       submissionDetails ─────┤
       questionData ──────────┤
                              ▼
                    Nunjucks render ──► files (+ assets)
                              │
                    [submission hook] ──► may rewrite the files
                              │
                    create commit (author.date = submission time)
                              │
                    ...repeat oldest → newest...
                              │
                    [post hook] ──► optional extra commit
                              │
                    ONE ref update  ◄── atomic push

Who it's for

Solution vaults — one README + code file per problem, with frontmatter ready for a static site generator.

Polyglot learners — keep every language you solved a problem in, each in its own file, filtered to the languages you care about.

Typst users — convert problem statements into a printable workbook and compile it in a post hook.

Anyone migrating — point commit.prefix at your old sync tool's commit header and inherit its watermark, so only new submissions are pulled.

Start with Getting started, then tune the output in Configuration and Templates. If something bites, check Limitations and the FAQ.