Datasets

Name: qipu data: San Francisco
Creator: qipu
Published: 2026-10-21
License: https://creativecommons.org/publicdomain/zero/1.0/

San Francisco’s public records as tables: what the Board of Supervisors decided since 1906, who voted how, and the places the decisions name. Every row links to the scanned page on the Internet Archive it was read from.

Releasev0.7.1

Published21 October 2026

LicenceCC0 1.0

Coverage1906–2026

Tables49, as Parquet and SQLite

Data dictionary Changelog Headline numbers and their SQL Cite this releaseA release never changes after it is published; corrections ship as a new version.

Start here

Six tables answer most questions. Each is one Parquet file; the same tables are in the SQLite file.

Table	What it holds	Rows	Parquet
Decisionsdecisions	Every item the Board acted on, 1906–2026, with date, outcome, a one-line summary and the scanned page.	718,093 rows	89.2 MB ↓
Votesx_roll_call_votes	How each supervisor voted in every printed roll call, 1906–1998.	2,084,507 rows	3.8 MB ↓
Supervisorspersons	The 294 supervisors, with their Wikidata ids.	294 rows	< 0.1 MB ↓
Placesplaces	Streets, blocks, addresses and named places, with stable ids and locations.	470,596 rows	19.9 MB ↓
Scanned pagespages	Every archive.org page the data cites. We link to the scans; we never copy them.	183,888 rows	6.8 MB ↓
Sourcessources	Every volume and report cited, with its archive.org collection.	6,377 rows	0.3 MB ↓

Download the whole release

File	What it is	Size	Format
`parquet/`Parquet, one file per table	The reference format. Read in DuckDB, Polars, pandas or R; no need to download first.	49 files · 1510 MB	Parquet
`qipu_sf.db`SQLite, San Francisco	The same tables in one file, privacy-filtered, plus ready-joined tables. Opens in any SQLite tool or Datasette.	6.44 GB	SQLite
`csv/`CSV extracts	Decisions, votes, roll calls, pages, places and sources, plus the facts by kind and decade (the list is in CSV_MANIFEST.json).	94 files · 1.8 GB	CSV, gzip
`SQL console`In your browser	Query the Parquet files here, without downloading anything.	no download	DuckDB

All 49 tables including working tables

Each table is one Parquet file, and a table of the same name in the SQLite file. Its page has the columns and a preview; the columns are also in the data dictionary.

Table	Description	Rows	Parquet

Board of Supervisors decisionsdecisions	One row per item the Board acted on, as printed in the Journal or the minutes. Some items were read more than once, so an item can have more than one row.	718,093 rows	89.2 MB ↓
Printed roll callsx_roll_calls	Every printed roll call ("Adopted by the following vote: Ayes: Supervisors ... Noes: ... Absent: ...") found in the Journal text, 1906-1999.	171,123 rows	16.1 MB ↓
Votes in printed roll callsx_roll_call_votes	One row per name in a roll call: the vote, the name as printed, and the official it was matched to and how.	2,084,507 rows	3.8 MB ↓
Ordinances, resolutions and billsinstruments	Ordinances, resolutions and bills by printed number.	72,224 rows	0.9 MB ↓
Numbers printed in decisionsdecision_numbers	Decision -> number(s) printed in its text.	88,876 rows	0.9 MB ↓

Claimsclaims	One fact per row (a thing, a place, a time), each with the page it comes from. Every source gives facts of the same shape.	8,367,105 rows	807.5 MB ↓
Claim supportclaim_support	The readings behind each fact: when several readings found the same statement, it is one fact with several supports.	6,594,525 rows	137.7 MB ↓
Claim reliabilityclaim_reliability	How firmly each fact is backed: verified (checked against the scan by a person, or stated by two documents), quoted (its words are on the cited page) …	8,367,105 rows	132.2 MB ↓

Placesplaces	Today's places with stable qipu ids.	470,596 rows	19.9 MB ↓
Place geometriesplace_geometries	Dated shapes of places (EPSG:4326): parcel and block outlines, address points, street lines, neighbourhoods and the 1906 burn zone.	459,992 rows	137.7 MB ↓
Place namesplace_names	Names of places, official and other. Today's names and search aliases for now; historic names will follow.	2,274 rows	< 0.1 MB ↓
Address resolutionaddress_resolution	Addresses as printed ("<number> <street>") matched to today's city address list. Only exact, unique matches.	805,748 rows	7.8 MB ↓
Personspersons	The officials in the data: members of the Board of Supervisors from the official roster, with Wikidata ids, and four printed in the Journals but …	294 rows	< 0.1 MB ↓
Person names as printedperson_names	Each name as printed in roll calls, with the years seen and the person it resolves to.	9,251 rows	< 0.1 MB ↓
Rolesroles	Supervisor terms (contiguous years served, from the roster).	318 rows	< 0.1 MB ↓
Organisationsorgs	Organisations (v1: the Board).	1 row	< 0.1 MB ↓
Place relationsplace_relations	How places relate, computed from geometry each build (plus declared links): an address within a block, a block within a neighbourhood, a street crosses …	1,284,690 rows	5.3 MB ↓
Block pointsblock_points	One point per assessor block (the map's block index), for fast lookups.	5,102 rows	< 0.1 MB ↓

Sourcessources	Every source cited. archive.org items carry their collections and Democracy's Library membership.	6,377 rows	0.3 MB ↓
Pagespages	Every archive.org page cited. We point at the scans; we never copy them.	183,888 rows	6.8 MB ↓
Runsruns	What made each table: rules, hand work, or a model run (model, prompt).	43 rows	< 0.1 MB ↓
Field scoresfield_scores	Measured accuracy of each reading, field by field, with the sample each was measured on (model checks and answer keys; no person has checked a sample yet).	36 rows	< 0.1 MB ↓
Identifier redirectsid_redirects	Old or merged identifiers and the identifier each now points to, so links keep working.	135,661 rows	1.5 MB ↓
Correctionscorrections	Corrections made by people, kept next to what was read. The site applies them at once; the files get them with the next release.	0 rows	< 0.1 MB ↓
OCR filesocr_files	The OCR files on archive.org behind each source item (the manifest the page texts were read from), so a reader can fetch and check them.	11,323 rows	0.6 MB ↓

Decisions, per runx_decisions	One row per Board item per reading, with the columns each reading filled.	748,894 rows	94.0 MB ↓
Map decision rowsx_site_rows	The decision rows shown on qipu's map, with the Board item each comes from.	62,917 rows	3.5 MB ↓
Redevelopment Agency appraisalsx_appraisals	The Redevelopment Agency's property files for the Western Addition A-2 project, one row per parcel page, matched to today's parcels.	3,189 rows	0.3 MB ↓
Democracy's Library place tagsx_dl_tags	Democracy's Library items about San Francisco, with the neighbourhoods, streets and places each is about (read by a language model).	1,131 rows	< 0.1 MB ↓
Mediamedia	Every photo, film, recording and document qipu's map shows, with its credit, licence and holder.	1,428 rows	0.1 MB ↓
Democracy's Library factsx_dl_facts	Facts read page by page from Democracy's Library documents about San Francisco (costs, homes, dates, places), each with its page and quote.	150,446 rows	11.9 MB ↓
Assessor real-property directoryx_realprop_dir	Extraction table: the Andrews Real Property Directory of San Francisco, 1943 and 1949 editions (archive.org scans), one row per printed address line: …	276,599 rows	4.0 MB ↓
Mentionsx_mentions	Democracy's Library items and the neighbourhoods and streets they mention (one row per mention; from the DL place tags, x_dl_tags).	915 rows	< 0.1 MB ↓

Claim propertiesproperties	The claim property vocabulary (spec/vocab/properties.csv).	85 rows	< 0.1 MB ↓
Place typesplace_types	The place type vocabulary (spec/vocab/place_types.csv): id segment, family, clickable buffer, salience, Wikidata class.	82 rows	< 0.1 MB ↓
Citiescities	City configuration (config/cities/*.json): namespace, currency, privacy embargo, address authority.	2 rows	< 0.1 MB ↓
Privacy rulesprivacy_rules	Names of private people are public after N years: sf 75, van 92.	2 rows	< 0.1 MB ↓
Eventsevents	Events (for now one: the 1906 earthquake and fire, with the area the fire burned).	1 row	< 0.1 MB ↓
Consumer price index, United Statescpi_us	Vocabulary: the US consumer price index by year (1982-84 = 100), used to state amounts in 2026 dollars (scripts/18_cpi.py). BLS CPI-U U.S. city …	227 rows	< 0.1 MB ↓
Consumer price index, Canadacpi_ca	Canadian consumer price index per year (and the latest month), used for 2026 dollars on Vancouver amounts: Statistics Canada BLS CPI-U U.S. city …	113 rows	< 0.1 MB ↓

Entity idsentity_ids	v0.7: the minted-id registry of persons and organisations (ids/persons.csv, ids/orgs.csv in the code repo, append-only). An id is minted once and never …	336 rows	< 0.1 MB ↓
Org mergesorg_merges	v0.6.2 stub (empty in the release): the merge log of organisation names. One row per printed name variant folded into an organisation, with the rule …	0 rows	< 0.1 MB ↓
Place mentionsplace_mentions	v0.7: every place string a statement was read with, one row per page and string; claims.place_mention_id points here. The input to place resolution, …	892,915 rows	17.1 MB ↓
Withdrawnwithdrawn	v0.7: record time. Every id that was in an earlier release and is gone from this one, with why and what replaced it; carried forward from release to …	54,652 rows	0.3 MB ↓
Blocks and census tractsx_block_tract	Which census tract each city block falls in, for each census map. From DataSF and the Census, not the Internet Archive.	20,389 rows	< 0.1 MB ↓
Registered business locationsx_business_locations	Registered business locations (DataSF) matched to today's address and block, with start and end years. Owner names are not included.	248,990 rows	6.5 MB ↓
Census tracts, 1980–2020x_census_tract	People, households and homes per census tract, 1980 to 2020, with median household income. US Census Bureau.	925 rows	< 0.1 MB ↓
Building permits by blockx_permits_block	Building permits per city block and kind (DataSF), filed 1982–2026: how many, and the first year of each kind.	28,708 rows	< 0.1 MB ↓
Outside identifiersxref	Outside identifiers (Assessor block and lot, DataSF address id, ordinance number, Wikidata, Internet Archive item and page, census tract) and the qipu …	1,320,115 rows	4.0 MB ↓

Documents and other files 13

File	What it is	Size	Format
`README.md`	What the release is and how to read it	11 KB	Markdown
`DATA_DICTIONARY.md`	Every column of every table	263 KB	Markdown
`DATASHEET.md`	How the data was made, its gaps and its uses	10 KB	Markdown
`CHANGELOG.md`	What changed in each release	32 KB	Markdown
`CITATION.cff`	How to cite the release	1 KB	CFF
`LICENSE`	CC0 1.0 for the facts; the scans keep their own rights	2 KB	Text
`NUMBERS.json`	The headline numbers, each with its SQL	4 KB	JSON
`manifest.json`	sha256 of every file, build commit, checks	18 MB	JSON
`release.json`	File list with rows and sizes	9 KB	JSON
`metadata.json`	Datasette metadata	10 KB	JSON
`showcase_queries.json`	Example queries for the six places searched in depth	8 KB	JSON
`qipu-data-scripts-v0.7.1.tar.gz`	The scripts that built the release	734 KB	tar.gz
`iiif/`	IIIF annotation pages: 109,079 files, listed in manifest.json	721 MB	IIIF JSON

Cite this release

qipu (2026). San Francisco public records, release v0.7.1 [data set]. 21 October 2026. https://data.qipu.org/usa/san-francisco/. CC0 1.0.

There is no DOI yet. Each release will get its own archive.org item.