HOA Knowledge Assistant
A retrieval-based assistant that answers residents' questions from a community association's own website content, with sources.
- Python
- WordPress REST API
- Embeddings
- Vector database
- RAG
Pipeline
- 01 WordPress API Pages, posts, documents
- 02 Chunk + embed Vector database
- 03 Retrieve Top matching passages
- 04 Answer Grounded, with sources
The problem
A homeowners association website holds years of rules, guidelines, meeting notes, and announcements. Residents can’t find the one paragraph that answers “Can I install a fence?” and end up emailing the board.
How it works
Content is pulled through the WordPress REST API, cleaned, split into passages, and embedded into a vector database. A question is embedded the same way, the closest passages are retrieved, and the model answers using only those passages, linking to the source pages.
I prototyped the knowledge base in NotebookLM first to check that the source content could answer real questions before building the pipeline.
How I test it
- A set of real resident questions with known answers and known source pages.
- Retrieval check: does the right page appear in the top results?
- Grounding check: is every claim in the answer supported by a retrieved passage?
- Refusal check: for questions the site doesn’t cover, the assistant should say so instead of guessing.