I Built a Serverless Stranger-Matching Tool with Zero Database — Here is the Architecture

Short summary

fata.uk is a serverless stranger-matching service using BGE-M3 embeddings to find emotional resonance with zero database or signup — the entire backend is one Cloudflare Worker. State and encrypted user data live in GitHub Issues and Cloudflare KV; after matching via LLM scoring, users communicate only via email so fata cannot read messages. The architecture combines semantic matching, multi-layer rate limiting, proof-of-work challenges, and survived a 3.49B-token LLM spam attack.

• •Serverless stranger-matching using embeddings and LLM scoring, zero database architecture
• •Encrypted state stored in GitHub Issues and Cloudflare KV; communication happens only via email
• •Multi-layer abuse prevention with PoW, rate limiting; includes postmortem of LLM spam incident

Generated with AI, which can make mistakes.