Code-Ledger

From the README

CodeLedger

Your DSA grind, on your GitHub. Automatically.

Zero-click commits · AI code review · Live analytics · Knowledge graph · Bulk LeetCode import

The Problem

You grind DSA every day. Your GitHub looks empty. Your solutions disappear into LeetCode's servers. When a recruiter asks to see your work, you have nothing to show — even though you've been coding for months.

CodeLedger fixes that. Install the extension, connect your GitHub, and every accepted solution gets committed automatically — code, problem statement, AI review, and all. Your profile becomes a living portfolio. Zero extra steps.

How It Works

The commit path and the review path are deliberately separate. A slow or failing AI provider never delays or blocks the commit. A finished review is written to IndexedDB and the problem is flagged pending; a maintenance alarm running every ten minutes rolls all pending reviews and metadata into one further atomic commit, rather than one commit per review.

What Gets Committed

One accepted submission produces one atomic commit through the GitHub Trees API — never a partial write, never one commit per file.

problems/two-sum/leetcode/
├── lc-two-sum.py        ← your code, exactly as submitted
└── README.md            ← statement, hints, runtime/memory, AI review

index.json               ← machine-readable history (rewritten every commit)
README.md                ← stats dashboard, regenerated in the same commit
index.html               ← GitHub Pages dashboard, same commit
.codeledger/
├── sync.json            ← portable settings backup
├── behaviour-bank.json  ← your AI skill definitions and insights
└── roadmaps.json        ← learning roadmaps

two-sum is the canonical ID. When the same problem exists on more than one platform, every platform's solution lands under the same canonical directory, so problems/two-sum/leetcode/ and problems/two-sum/geeksforgeeks/ sit side by side. If no canonical mapping is known, the layout falls back to problems/lc-two-sum/.

Install

Loading unpacked from source:

Chromium   chrome://extensions  →  Developer mode ON  →  Load unpacked  →  select dist/chromium
Firefox    about:debugging      →  Load Temporary Add-on  →  select dist/firefox/manifest.json

Run npm run build first for either target. src/ holds two manifests — manifest-chromium.json and manifest-firefox.json — and no manifest.json; picking one and writing it out under that name is what the build does, so src/ cannot be loaded unpacked directly.

Setup (2 minutes)

1. Click the CodeLedger icon in your toolbar.
2. Click Connect GitHub. The OAuth exchange runs through the Cloudflare Worker at codeledger.vkrishna04.me, which swaps the authorisation code for a token and returns it to the extension. The token is stored in your browser and never on the worker.
3. Pick a repo name (e.g. my-dsa-solutions). CodeLedger creates it and writes the first commit itself.
4. Solve a problem. Check your GitHub.

CodeLedger requests public_repo and workflow by default — enough to create a public repository, push to it, and write the GitHub Pages workflow. Choose the broader repo scope during authorisation if you want your solutions in a private repository.

Optional: add an AI provider key under Settings → AI to turn on code review.

Supported Platforms

NeetCode and takeuforward are single-page apps that judge over fetch and never render a durable verdict node, so their handlers read the judge's own response instead of the DOM. On takeuforward only the TUF+ judge reports a verdict; without a subscription the handler still marks off sheet rows you have already solved elsewhere, but has nothing to commit.

Each platform is a self-contained directory under src/handlers/platforms/. Adding another one requires no changes to the core — see the handler contract.

Supported AI Providers

Keys are stored in your browser under ai.keys and are sent only to the provider you selected. If a provider fails, the queue retries and can fall back to another configured provider.

What You Get

Architecture

No bundler, no transpiler. Pure ES6 modules; Preact and htm come from src/vendor/preact-bundle.js, a local bundle generated by npm run vendor:preact and committed to the repo, so nothing is fetched from a CDN at runtime. Tailwind is compiled ahead of time into a single stylesheet.

Two rules hold everywhere:

• Every extension API call goes through src/lib/browser-compat.js. Nothing else touches chrome.* or browser.*.
• Every log goes through createDebugger() from src/lib/debug.js, which preserves the caller's file and line in DevTools.

See docs/architecture/system-architecture.md for the full component breakdown.

The Worker

A single Cloudflare Worker (Hono) does two jobs: it relays the GitHub OAuth code exchange, and it serves the static landing page.

The route contract is specified in docs/OPENAPI.yaml and validated in CI.

Privacy

Your solutions, your tokens and your AI keys stay in your browser and go only to the services you chose. There is no account and no server-side copy of your data.

One optional exception: Anonymous Usage Stats, off unless you turn it on in Settings → Advanced. When enabled, each committed solve sends a counter hit containing the platform name and extension version — nothing else, no identifier of any kind. Full detail in PRIVACY.md.

For Developers

npm install
npm run build:css        # Tailwind → src/ui/styles/compiled.css
npm run build            # CSS + per-target dist/
npm run lint             # dev/typecheck.js — tsc --checkJs over every file, fails on undefined names
npm run format           # prettier over src, dev, worker
npm test                 # node:test — extension + worker suites
npm run validate:openapi # worker routes must match the spec
npm run release          # validate → build → zip → commit → tag → push
npm run release -- --dry-run

Automation

Seven workflows in .github/workflows/. The four that run Node pin Node 24; the three store-publish workflows need none.

All three store workflows skip pre-release tags — any tag containing a hyphen.

Adding a platform handler

1. Create src/handlers/platforms/{name}/index.js extending BasePlatformHandler.
2. Add dom-selectors.js (with a DOMAINS export) and page-detector.js alongside it.
3. Add the hostname branch in src/content/handler-loader.js.
4. Run npm run domains:update to regenerate host_permissions in both manifests.

The full contract is in docs/guides/development/adding-platform-handler.md. LeetCode's handler is the reference implementation.

Adding an AI provider

1. Create src/handlers/ai/{name}/index.js extending BaseAIHandler, plus model-fetcher.js.
2. Add the provider to CONSTANTS.AI_PROVIDERS in src/core/constants.js.
3. Register its settings schema in src/handlers/init.js and wire it into ModelSelector.js.

Self-hosting the worker

1. Register a GitHub OAuth App (not a GitHub App — GitHub Apps ignore the scope parameter and issue user-to-server tokens that cannot create repositories). Callback URL: https://<your-worker>/api/auth/github/callback.
2. Copy worker/wrangler.toml.example to worker/wrangler.toml, then set the secrets it lists with npx wrangler secret put <NAME>.
3. Point CONSTANTS.URLS.AUTH_WORKER in src/core/constants.js at your worker.
4. npm run deploy:worker.

Contributing

Read .github/CONTRIBUTING.md before opening a PR. Open tickets live in docs/reference/backlog.md.

Before pushing: npm run lint && npm run format:check && npm test.

Documentation

• Changelog: docs/CHANGELOG.md
• Architecture: docs/architecture/system-architecture.md
• Queues and orchestration: docs/queues/orchestration.md
• Testing guide: docs/TESTING_GUIDE.md
• OpenAPI contract: docs/OPENAPI.yaml
• Adding a platform: docs/guides/development/adding-platform-handler.md
• Graphify knowledge-graph workflow: docs/guides/development/graphify-workflow.md
• Security policy: .github/SECURITY.md

codeledger.vkrishna04.me · Apache 2.0

Fork it. Self-host it. Own your data.