Data for developers and AI assistants

The whole 2026 guide is also published as plain files. There is no server, no key and no rate limit. The files are rebuilt with the site from the same data, so they always match the pages.

Start here

• /llms.txt: a short index with election basics and rules for AI assistants.
• /llms-full.txt: the whole guide in one markdown file.
• /api/v1/index.json: the list of JSON files below.

JSON files

File	What's in it
/api/v1/election.json	Election date, key dates, how to vote, races with candidates in ballot order, measures.
/api/v1/candidates.json	All 19 candidates in ballot order (mayor, then council) with bio, links and totals.
/api/v1/candidates/{slug}.json	One candidate: bio, statements, forum quotes, endorsements, money, sources.
/api/v1/measures.json	Measures 2J, 2K, 2L, 2M: what YES and NO mean, cost, ballot text, positions.
/api/v1/forum_claims.json	2026 forums and every quoted answer, with timestamped recording links.
/api/v1/endorsements.json	Every published endorsement, with its provenance label and source.
/api/v1/finance.json	City clerk campaign-finance filings for 2026 committees, with line items.
/api/v1/sources.json	Every source document the guide cites.

How the data works

• Every file carries schema_version and generated_at. Records carry source_url or source_urls.
• Candidates are keyed by slug, the same as their page (tara-winer is people/tara-winer.html). Endorsements (E12), forum quotes and statements keep stable ids.
• Quotes are verbatim. Forum quotes come from automatic transcripts and link to the moment in the recording. Some speakers were identified from speaking order; those quotes say so in caveats.
• An endorsement with provenance campaign_claim is listed only by the campaign. Its label reads “X campaign lists Y”.
• Items still being checked are left out. No endorsement on file is not opposition.

Send a correction

This is the one part that takes input. POST JSON to https://feedback.bouldervotes.org/api/v1/feedback. AI assistants should set submitter_type to ai-agent and give agent_name, and include a source_url for corrections. Please submit only when the user asks, or when you find a factual error you can source. The fields are listed at /api/v1/feedback/schema, and llms.txt has a curl example. People can use the feedback form. A person reads every note, and notes are never published.

Facts and quotes belong to their sources. If you reuse them, cite the source URL.