What this is

A federal Australian sister to House (UK), built for Guardian Australia journalists who need to triage Hansard and Senate Estimates faster than the official tools allow. Three surfaces: Search across spoken contributions, questions on notice and ministerial statements; Deep Dive for plotting a term over time as a party-stacked timeline; Estimates for searching twice-yearly Senate hearings by round, committee, witness and senator. The third is the headline AU surface — UK House has no equivalent because the UK Parliament has no equivalent.

Use cases that drove the shape: a defence correspondent reading everything said about AUKUS in a single sitting fortnight; a political reporter tracking how a junior minister has answered Estimates questions over a year; an investigations editor scanning every Estimates exchange about robodebt or sports rorts; a tech reporter pulling every Estimates appearance from a specific department secretary.

How it’s made

Vanilla JavaScript on GitHub Pages, no proxy, no live API. UK House leans on UK Parliament’s live, CORS-friendly endpoints; aph.gov.au doesn’t expose an equivalent. So House AU inverts the architecture: a nightly GitHub Action walks the latest sitting days, fetches contribution JSON via parliament’s own internal transcript endpoint, parses it into speaker-attributed rows, and commits a tiered set of static shards back to the repo — per-month for the live two years, gzipped per-quarter for the older archive. The browser fetches the manifest, picks the shards overlapping the user’s date range, and searches them client-side. The cost is a ~24-hour lag, which matters less than it sounds because Hansard itself runs about a day behind the chamber.

Senate Estimates runs through a parallel pipeline because the data shape is different: each fragment covers an entire department’s grilling for one day rather than a single speaker turn, and the editorially useful unit is the section (department) rather than the turn. Same harvester architecture, different parser, separate per-round shards.

Built with Claude Code, internal newsroom tool, currently hosted on my personal GitHub. The Parliament of Australia publishes its material under CC-BY-NC-ND 4.0 — non-commercial, no derivatives. Counsel are relaxed about the framing because this is a research surface for journalists, not a public republication of a derivative work. The expectation is it moves to the Guardian stack behind authentication once it’s doing useful work.

What it doesn’t do

It doesn’t mirror UK House’s Committees and Transcripts surfaces. UK Parliament’s Committees API exposes oral evidence transcripts as searchable text; aph.gov.au doesn’t, and the underlying ParlInfo XML sits behind an Azure WAF that blocks automated requests. Building it for AU would have meant a third parallel harvester pipeline fighting the source at every step, with most of the editorially interesting committee work already covered by Senate Estimates. Cut on purpose.

Also not in here: alerts, saved searches, cross-state coverage, the New Zealand sister tool somebody will eventually ask about. Federal first; widen the scope only if the narrow version actually gets used.