QED — computational knowledge engine
QED is a Wolfram Alpha-style computational knowledge engine: ask a question in natural language and QED returns a structured, deterministically computed answer.
Why QED?
Large language models (LLMs) are excellent at understanding questions but unreliable at computing and stating facts. QED therefore works the opposite way from a chatbot:
- zero hallucination — the LLM never computes math or generates factual answers;
- deterministic execution — all math, physics and fact retrieval go through dedicated tools and verified data sources;
- LLM as planner — the language model only orchestrates backend tools for complex queries;
- assumptions labelled — Fermi-style estimates are always clearly flagged, never silently presented as fact.
What does QED know?
Nine knowledge domains: arithmetic and unit conversion, symbolic math, plotting, chemistry, weather, geography and facts (Wikidata), nutrition, linguistics (including the Croatian HJP dictionary) and date arithmetic.
Example queries:
| Query | Answer |
|---|---|
30% of 8 miles |
2.4 miles |
balance H2 + O2 -> H2O |
2H₂ + O₂ → 2H₂O |
weather in Zagreb tomorrow |
real forecast (Open-Meteo) |
capital of Croatia |
Zagreb (Wikidata) |
Integrate[x^2, x] |
x³/3 |
Queries work in five languages: English, German, French, Spanish and Croatian.
Documentation
Detailed documentation is available in the QED documentation section: architecture, quick start, API and configuration.