About CivilAI
Municipal code research built like an engineering desk.
CivilAI is a focused research workspace for people who need to interpret local code without losing the evidence trail. It combines PDF parsing, structured storage, semantic retrieval, and saved chat history so ordinance questions become easier to check, revisit, and explain.
Why it exists
Because project decisions should not depend on PDF scavenger hunts.
Civil engineers, planners, consultants, attorneys, and municipal staff often need fast answers from long ordinance documents. Standard search can miss context, and general AI can drift. CivilAI is designed to keep the model close to the actual code sections and make the answer easier to inspect.
1 · Upload
Add ordinance PDFs.
Uploaded codes are stored by the backend and immediately sent through the parsing pipeline so they can become searchable workspace material.
2 · Structure
Turn pages into code hierarchy.
The parser extracts chapters, sections, subsections, summaries, source filenames, and navigation metadata. That structure is saved in SQLite and mirrored into Chroma.
3 · Ask
Query the code like a workspace.
The chat retrieves exact references, keyword matches, and semantic matches before the model drafts an answer from the narrowed evidence.
What happens behind the scenes
A structured retrieval stack, not a generic chatbot wrapper.
The backend saves PDFs in the data folder and parses them into normalized document, chapter, section, and subsection records.
The retrieval toolkit builds a navigation map, resolves jurisdiction aliases, and checks for exact section or subsection references in the user question.
Chroma and keyword search find supporting evidence. The system merges, reranks, and returns a compact context package to the language model.
The frontend displays the answer with source sections, jurisdiction focus, accuracy estimate, and saved chat continuity when the user is signed in.
Who it helps
For teams that need code answers they can defend.
Engineering teams
Quickly check requirements, summarize sections, and keep a citation trail attached to recurring project questions.
Municipal staff
Make local code easier to navigate for repeat questions, internal reviews, and resident-facing explanations.
Consultants and reviewers
Move faster through unfamiliar jurisdictions while keeping the final answer tied to the source document.
Start working
Bring a real ordinance question and test the trail.
The fastest way to understand CivilAI is to ask the kind of question that usually slows your review down. Try a section summary, a requirement lookup, or a jurisdiction specific code question.