@content-vigilante/core

Core engine for Content Vigilante — an open-source brand voice auditor.

Pre-alpha — 0.0.0 · This is a placeholder publish to reserve the npm scope. Full v0.1 ships when the engine is complete (target: ~3 weeks).

What this will do

import { audit, createAnthropicProvider } from '@content-vigilante/core';

const result = await audit({
  content: 'We are pleased to announce our new synergistic platform...',
  contentLanguage: 'en',
  guide: mailchimpGuide,
  llm: createAnthropicProvider({ apiKey: process.env.ANTHROPIC_API_KEY }),
});

console.log(result.score);   // 64
console.log(result.issues);  // [{ line, type, severity, text, rule, suggestion }, ...]

Roadmap

• [x] LLM provider abstraction (Anthropic, OpenAI, Ollama)
• [x] EmbeddingProvider interface
• [x] Tone judge skeleton with multi-shot LLM grading
• [ ] Brand guide ingestion (PDF + markdown + JSON) — week 1 day 2
• [ ] sqlite-vec embedding store — week 1 day 3
• [ ] Vocabulary judge (deterministic + LLM hybrid) — week 1 day 4
• [ ] Structure judge (deterministic) — week 1 day 4
• [ ] URL content extractor — week 2

See the main README for the full architecture and roadmap.

License

MIT for all code. The bundled mailchimp.json reference dataset is CC BY-NC 4.0 — see NOTICE.