⌘K
Agentic RAG Pipeline
Upload Documents
Ingest filings, permits, bulletins, and reports. Documents are parsed, chunked, embedded, and indexed for the AI Analyst.
Parse
PDF / DOCX / XLSX / HTML
Chunk
Semantic + layout-aware
Embed
1536-d vectors
Index
Hybrid BM25 + vector
Ingest into knowledge base
2/3 indexed
Drop files here or click to browse
PDF, DOCX, XLSX, CSV, TXT, MD, HTML, PNG, JPG — up to 100 MB per file
PDF
DOCX
XLSX
CSV
HTML
Images
OCR scanned pages
Vision model for image PDFs
Auto-tag entities
Geo, parcel, policy refs
Knowledge base
3
Docs
2,256
Chunks
504.9K
Tokens
Governance
- PII redacted on ingest
- Source URLs preserved as citations
- Re-embedding on policy change
Retriever
Hybrid BM25 + dense (top-k 8) → reranker → tool-bound agent. Average recall @ k=5: 0.91
Ingested documents
HCAD_2026_Appraisal_Roll.pdf
Indexed
17.5 MB·Appraisal·1842 chunks·412.3K tokens
Houston_Permit_Activity_Q2.xlsx
Indexed
2.0 MB·Permits·318 chunks·71.2K tokens
TX_Insurance_Bulletin_B-0042.pdf
Embedding
957.0 KB·Policy·96 chunks·21.4K tokens