Everything people ask us, answered straight.

Public signup is closed. This site presents Govafy as a project by Abraham Xiong. The Project Details page documents its architecture, technology choices, and lessons.

How government data, company knowledge, and AI-assisted workflows can come together in one workspace for opportunity research, capture, and proposal work.

The Project Details page covers TypeScript, data ingestion, PostgreSQL, OCR, RAG, the application stack, and the distinction between the application and this website.

Retrieval-augmented generation finds relevant source material and provides it as context for an AI response. Govafy combines keyword and vector retrieval where embedding coverage permits, with source selection scoped to the organization and project.

The document pipeline attempts digital extraction first. Scanned PDFs can use local Tesseract OCR, with Google Document AI as a fallback. Security checks and extraction quality affect whether a document can enter retrieval.

No. Structured award records and document text follow different paths. Ingestion, extraction, chunking, and embedding are separate stages; the award-record count is not a count of embedded documents.

Yes. Source references support review, but a person still needs to check factual claims, past performance, requirements, and the proposed response before using the work.