Municipal Code Review

What this is

Inspired by BetterEU, this project runs Solano County municipal codes through LLMs with transparent, published prompts. The goal isn’t “keep or delete” – it’s explain, flag, compare:

Plain-language summaries – Each code section gets an LLM-generated summary a non-lawyer can understand.
Flags – Identify sections that are outdated, redundant, internally contradictory, or unusually restrictive compared to peer jurisdictions.
Cross-jurisdiction comparison – Highlight where Solano cities differ on the same topic (ADU regulations, business licensing, noise ordinances).
Prompt transparency – The exact prompt and code are published so residents can evaluate the AI’s reasoning.

Current status

Proposed. The concept is designed and documented. Municipal code sources have been identified for Vacaville, Fairfield, Solano County, Dixon, and Vallejo. No code has been scraped or analyzed yet.

Phasing

Phase 1: Capture – Scrape, parse, and normalize municipal code into structured data.
Phase 2: First analysis – Plain-language summaries for one jurisdiction as proof of concept.
Phase 3: Additional analyses – Flags and cross-jurisdiction comparisons as independent passes over the same data.
Phase 4: Public feedback – Residents can submit corrections and context on specific sections, with visible status tracking.

What we need

Platform identification. Which hosting platform does each jurisdiction use? (Municode, American Legal, Sterling Codifiers, custom?) This affects the scraping strategy.
Legal expertise. If you’re a local attorney or planning professional who can validate LLM summaries against the actual legal meaning, your review is especially valuable.
Technical contributors. Python, scraping, LLM API experience. The architecture is designed but not built.
Resident testers. When summaries are generated, we need people who’ve actually dealt with the code (building permits, business licenses, noise complaints) to say whether the summaries match reality.

Design principles

Section is the unit of analysis. Each code section is one LLM call. This maps to how the code is structured and how amendments target changes.
Cache aggressively. Code doesn’t change often. LLM outputs are stored keyed to a hash of the input. Only re-analyze when content changes.
Every piece of feedback has a visible status. Submitted, acknowledged, categorized, responded to. No black holes.
Feedback is public. Other residents can see what’s been submitted and how it was handled.

How to contribute

See the contributing guide for how to get involved with OpenSolano projects, including how to submit data corrections and improvements via GitHub.