Official first
Portland.gov records are the factual foundation. This guide distinguishes official facts from normalization, calculation, and Portland Civic Lab analysis.
Public method
Inspect the method, not just the result.
This guide is designed so every signal can be traced to inputs, rules, and official sources.
Unknown is valid
Missing public information is labeled as unavailable or unable to determine. Absence is never silently converted into certainty.
Readiness, not approval
Decision readiness checks whether decision-relevant information is present. It never recommends a yes or no vote.
Append-only history
Every observed page and document version is preserved by content hash so prior public states can be reconstructed.
New-government boundary
The complete-record view begins January 1, 2025. It reconciles both current and completed indexes instead of treating the active index as the whole legislative record.
Votes stay contextual
Final actions, amendments, committee recommendations, and procedural motions retain separate scopes. Councilor position patterns and colleague comparisons use verified final actions only; a vote against an amendment is never relabeled as a vote against the underlying ordinance.
Ambiguity is quarantined
The vote parser checks declared tallies against unique listed names and rejects contradictory position assignments. Ambiguous agenda vote blocks remain in the audit record but are withheld from public totals and Councilor analysis until reviewed.
Positions require the full record
Councilor position guides review every available current-term named vote, motion, sponsorship, and attributed meeting-transcript section, not a hand-picked excerpt sample. The published synthesis separates outcomes, policy instruments, stated reasons, exceptions, and unanswered questions.
Local, version-controlled synthesis
The project exports a frozen evidence bundle and uses a Codex editorial task to organize it by policy topic. The reviewed result is checked into Git as data. The deployed application makes no model calls and requires no AI provider key or usage billing.
Claims stay accountable
The local validator rejects a stale evidence hash, unknown or mismatched source IDs, motion-level votes cited as final-action position evidence, non-verbatim quotations, missing topic analyses, duplicate topics, and colleague statistics that do not match the divided final-action record. Synthesis may not assign motive, private belief, ideology, virtue, effectiveness, or electoral intent.
Monthly, versioned refreshes
Once a month, an editor re-exports each Councilor's evidence bundle in a Codex task. A changed hash requires a new review before the replacement can be checked in and deployed; the site never silently rewrites a published guide.
Major topic selection
A topic dossier must score at least 8 of 12 and must demonstrate both a continuing public record and meaningful public consequence. Complexity, disagreement, unresolved relevance, and the need to connect multiple records supply the remaining score. Popularity alone never qualifies a story.
Coverage is measurable
Introducer, official text, attendance, agenda linkage, motion, amendment, and vote coverage are published on the data-health page.
Specification coverage
See measured coverageQuestions this guide can answer.
What is Council deciding, where is it in the process, and what happens next?
AvailableMatter overview, current pipeline, appearances, final disposition, and append-only timeline.
Who introduced it, who moved it, and who seconded it?
AvailableNormalized sponsor records and agenda-derived motion attribution link to person profiles.
How did each member vote on each motion and final action?
Available with scopeMotion-level votes come from agenda pages; the official vote register supplies final-action positions. Ambiguous parsed blocks are visibly counted in data health and withheld until reviewed.
What changed, including amendments?
Available with scopePublished amendment motions and official source versions are preserved. Unpublished drafting changes cannot be reconstructed.
What does it cost and how is it funded?
Available when publishedOfficial impact statements plus visibly labeled keyword-extracted amounts; no amount is invented.
What alternatives, assumptions, risks, dependencies, evidence, and outcome measures appear?
Available when statedDeterministic, source-linked passages are grouped by category; missing categories say not found.
What did Council direct, promise, or require after the vote?
Available when statedOfficial directions and commitments are extracted verbatim and connected to the matter.
What similar work happened before?
Available since Jan. 2025Full-record search covers active and completed matters in the new-government period.
What is each councilor’s introduced-matter, motion, vote, and attendance record?
AvailablePerson profiles aggregate only named official records.
What has each councilor directly said and done on major policy topics?
Available with scopeCurrent-term guides synthesize all available named votes, motions, sponsorships, and attributed meeting speech, with source-linked claims, counterevidence, colleague comparisons, and explicit unknowns. The corpus does not yet claim to capture every public appearance or written communication outside these sources.
What questions did members ask and what answers did staff give?
Available with scopeAttributed automatic meeting transcripts support source-linked analysis and playback. Transcript wording can be imperfect, so direct quotes remain labeled and are treated as a finding aid rather than a substitute for the recorded meeting.
Did implementation produce the promised outcome?
Source-limitedDirections are tracked, but verified outcome monitoring requires later reports or operational integrations.