Skip to content
Solo builder, production product

Built by an engineer,
for engineers

The story behind Wauldo — why verified AI answers matter, and why one person built the whole thing in Rust.

Wauldo was born from a simple frustration: every RAG system I built hallucinated. LLMs would confidently cite sources that didn't exist, invent numbers, and present fiction as fact. My users deserved better.

So I built Wauldo — a RAG API where every answer is verified against source documents before it reaches your users. If the evidence is weak, we say so. If we can't verify it, we don't answer.

Today, Wauldo serves verified answers with 100% retrieval accuracy and 0% hallucination rate on production benchmarks. It's a one-person operation, and I intend to keep it that way — small team, focused product, no bloat.

By the numbers

What one engineer and a lot of Rust can do.

115k+
Lines of Rust
2,160+
Tests
59/59
Security findings fixed
17
Workspace crates
7
LLM providers
3
SDKs (Python, TS, Rust)

Philosophy

The principles that shape every decision.

Correctness over speed

We'd rather return "I don't know" than a wrong answer. Every response is verified against source documents before delivery.

Transparency by default

Every response includes confidence score, source citations, and model used. No black boxes, no hidden decisions.

No vendor lock-in

OpenAI-compatible API. Switch from any provider in 2 lines of code. Your data stays yours.

Tech stack

Production-grade infrastructure, no shortcuts.

Core Rust — async, type-safe, zero-cost abstractions
API Axum + Tokio — high-performance async HTTP
RAG BM25 + dense vectors + BGE reranking — hybrid retrieval
LLM OpenRouter — 5+ providers with automatic fallback
Infra Fly.io, Redis, FastEmbed — edge deployment
Quality Prometheus metrics, fact-checking pipeline, audit trail

Want to learn more?

Read the documentation, try the live demo, or start building with a free API key.