Guaraci: A Reproducible Orchestrator for Direct Extraction of Brazilian Public-Health and Environmental Data

Brazilian public data relevant to health and the environment are dispersed across heterogeneous official systems - FTP servers (DATASUS), REST APIs (OpenDataSUS/DEMAS), and web portals (gov.br) - and published in legacy, often compressed formats (e.g., DBC/DBF) with schemas that change across years and topics, which makes integrated, reproducible reuse costly. Guaraci is an open-source (MIT), Docker-first orchestrator that acquires such data directly from the primary official sources, decodes legacy formats, harmonizes heterogeneous schemas, records full provenance, and delivers the result to a downstream data lake. It integrates 90+ sources spanning public health (DATASUS microdata such as hospital and mortality information system; OpenDataSUS arbovirus and immunization data; gov.br sanitation indicators) and the environment (NASA POWER climate series, FIRMS active-fire detections and GPM IMERG precipitation), behind a single asynchronous job engine exposed through a command-line interface and a REST API. Data are decoded to records with pyreaddbc/dbfread, harmonized by taking the union of all historical columns (null-padding missing fields) and exported to CSV, Parquet and SQLite with a machine-readable file capturing source, request filters and output paths. By abstracting unstable FTP servers, paginated APIs and legacy encodings behind one reproducible mechanism, Guaraci lowers the technical barrier to Brazilian public data and continuously supplies analysis-ready inputs for spatio-temporal epidemiological modelling, machine-learning applications and cross-source studies.