This is an interactive way to interrogate the GDA public‑data lake —
779 datasets of official, published GB electricity‑grid data — in plain English.
You ask a question; it writes the query, runs it against the real data, and hands you the answer
together with the data itself and the means to verify it. Unlike the TSGB Expert (which reasons
over one author’s analysis), the answers here are computed directly from official source data.
What it is
Natural‑language access to the lake — operator, network and regulator published data (demand,
generation, system inertia, frequency, balancing and constraint costs, carbon, weather, capacity registers,
and much more), 779 datasets in all, described in a machine‑readable catalogue.
How it works, step by step
- You ask a question in plain English.
- The language model reads the catalogue of all 779 datasets, decides which are relevant,
and writes a single read‑only SQL query (DuckDB) over them.
- That query runs in a locked‑down sandbox: read‑only, no network, no file access
beyond the named datasets, resource‑capped, one
SELECT only — it cannot change or
exfiltrate anything.
- You get a ZIP containing: the plain‑English answer, the exact query that
ran, the result data (CSV + Parquet), and validation instructions
so anyone can reproduce and check it independently.
Where to be careful — the data is official, the query is the model’s
- The underlying data is trustworthy — official published grid data, not invented or
interpreted.
- But the model chooses which datasets to use and how to compute the answer. It can pick the
wrong dataset, misread your intent, mishandle a time zone or unit, or aggregate in a way you didn’t
mean — and it will present the result just as confidently either way.
- That is exactly why every answer ships with the query and the validation instructions. Read
them. Check the query actually asks what you asked. The tool is built so you never have to take its
word for it.
What it can — and can’t — answer
Only what the 779 datasets contain: the GB electricity system. Ask it about rainfall in the Sahara in 1850 and
it will (correctly) tell you it can’t, and explain why from the catalogue. It grounds every answer in what
is actually in the lake, and refuses what isn’t.
It is still an LLM
It can make mistakes in reasoning or in SQL. Treat the returned data as authoritative (it is
official); treat the interpretation — which dataset, how it was computed — as
something to verify with the query and validation notes provided.
In short: a genuinely powerful, honest way to ask the official UK grid data questions without knowing SQL —
the data is real and sourced, the query is transparent and checkable, and the model’s interpretation
is yours to verify. Use the validation; don’t skip it.