Llmdoc – annotate codebase with LLM summaries only re-scan what changed
Smart incremental summaries, but chat-with-codebase tools like Cursor and Continue already solve this.
Scan your codebase for off-brand UI copy. Get a score. Share it.
Extracts user-facing strings from JS/TS, templates, Markdown, and many backend formats, then runs each snippet through OpenAI or Anthropic to score tone against selectable voice templates (Professional/Casual/Technical) and report file+line hits. BYO API key and optional score-sharing keeps raw text local by default — smart for privacy — though I'd like to see CI hooks, rule customization, and threshold tuning to tame noisy LLM judgments.
Frontend and full‑stack developers, product/UX writers, and localization/content teams
built a cli that scans your codebase for off-brand product copy. think eslint but for your brand voice.
`npx brandlint`
it extracts user-facing strings (jsx text, error messages, placeholders, i18n), checks them against a voice template (professional, casual, technical), and reports issues with file and line number.
works with anthropic or openai keys. nothing leaves your machine except the score summary if you choose to share it :)
supports ts/tsx, vue, svelte, html, json, yaml, markdown, php, python, and more.
would love feedback. what brand voice rules do you usually enforce in your products, if any?
Smart incremental summaries, but chat-with-codebase tools like Cursor and Continue already solve this.
Tree-sitter plus LLM hybrid finds auth logic Semgrep rules miss.
Validates llms.txt and AI robot rules before AI crawlers ignore your content.
Purpose-built LLM security linter covers OWASP Top 10, but static analysis has inherent blind spots.
Catches missing env vars and even warns when API keys expire.
First compliance scanner for India's DPDP Act with concrete violation detection, not just checklist theater.