Purpose
The local brain helps visitors and operators understand staffing models, route requests, collect better intake, use safe government-contracting language, and find the correct page or form.
Architecture
The brain loads brain/brain-corpus.json, then uses JavaScript keyword scoring and rule-based routing to answer questions. It stores optional local notes in browser localStorage.
Why this is useful
It gives the site a smart-feeling assistant without calling paid AI providers. It also protects the company from risky claims by repeating safe-claim warnings when government, jobs, proof, or certifications are discussed.
Limitations
It is not a real LLM. It cannot reason deeply, inspect uploaded files, access private systems, or verify current government registrations. For that, connect a server-backed model or authenticated tools.
Upgrade path
Keep the local brain as fallback, then add a backend function that calls a GPU-hosted model, Ollama, vLLM, Cloudflare Worker, or another approved inference endpoint. Add authenticated database memory when upstream auth is ready.