Clermont, Florida
- Documents harvested
- 24
- Portal
- docs.clermontfl.org
- Platform
- Laserfiche
Agenda packets via CivicClerk; minutes via Laserfiche.
Every reading the observatory publishes lists its sources in line. Every claim has a citation chain. The trail is the receipt — and it is verifiable from this page back to the original PDF or video recording on the city's own portal.
Source documents come from each city's own portal. Some cities use municipal CMS platforms (CivicClerk, AgendaCenter); others publish PDFs directly under municipal subdirectories. The harvest pipeline reads where the city posts.
Agenda packets via CivicClerk; minutes via Laserfiche.
PDF agenda packets and minutes posted directly under municipal subdirectory.
Granicus-family agenda center; standardized municipal CMS.
Each meeting bundles agenda + minutes + supporting attachments.
The harvest output is one file per meeting, with a stable frontmatter contract for jurisdiction, board, date, and items. The shape is what allows the lens projection to read across cities without losing per-city texture. The text is plain Markdown; the metadata is structured. Every file is committable, diffable, and version-controlled.
The repository where the harvest lives is the NLAA — the sibling project at ../zoning-signal/. The observatory you are reading is the publishing surface; the NLAA is the harvest and synthesis surface. Cross-reference is via symlink, never duplication.
Validation is human. Before any place dossier or signal brief ships, the citation chain is walked end-to-end — back to the source PDF, back to the video recording where one exists. Items are checked for vote tally, public comment, and outcome. When a citation breaks (URL rot, document withdrawn), the dossier holds and is reviewed; when a citation refines (a staff report supersedes a prior version), the dossier moves with it.
The source trail at the bottom of every reading is the visible receipt. The provenance graph in the underlying content is the structural one — six document types per reading, ordered, with dates and source IDs that resolve back to this page.
AI agents reading the observatory get the same provenance. Every MCP tool response includes the citation chain. Every projection preserves its sources. The discovery surface is /.well-known/mcp.json; the long-form ingest substrate is /llms-full.txt; the index is /llms.txt.
Read the methodology for how the substrate is read. See a place dossier for what the discipline produces.