AI-powered research and drafting assistant for academic papers, reports, and essays.

Humanoid acts as a multi-stage background agent that runs async pipeline stages (Research → Outline → Draft → Verify → Style → Format) and produces submission-ready documents with fully verified citations.

It is designed with a strictly minimal UI (black & white achromatic flat design) to enable zero-distraction focus, inspired by modern technical documentation.

• Frontend: Next.js (TypeScript, React 19)
• API: FastAPI (Python 3.12)
• Database: Supabase / PostgreSQL (with pgvector for embeddings and Row-Level Security for strict per-user isolation)
• Job Orchestration: Redis + arq (async worker framework)
• Providers: Gemini Flash (primary) + resilient fallback pool with backoff and tier-based routing
• Ingestion & Render Services: Fully isolated Docker containers processing untrusted inputs (PDFs, PPTs) and orchestrating headless LibreOffice

• Strict RLS: Every database table is locked behind Row-Level Security scoped strictly to the current auth.uid().
• Zero Hallucination Generation: RAG-constrained drafting means every claim requires a citation. Missing citations are explicitly flagged.
• No Detector Evading Design: This product does not optimize text to evade AI detection. It prioritizes clarity and truthfulness.

1. Clone this repository.
2. Duplicate .env.example to .env and fill in your Supabase credentials.
3. Start the entire application using Docker Compose:

docker compose up

MIT License