The file storage system. Holds all raw, cleaned and summarised research data organised in three buckets — Bronze, Silver and Gold. Works like an S3-compatible hard drive inside the server.

The reporting interface. Institutions, researchers, donors and administrators view KPI dashboards, satisfaction scores, geographic coverage and data quality reports here.

The ingestion bridge. When a researcher submits data on Dacoris, this service polls every 5 minutes, downloads the file, converts it to Parquet and stores it in the Bronze bucket.

The transformation engine. Reads Bronze data, builds 23 models across Silver and Gold layers — 2 survey-specific Silver tables and 21 Gold KPI models — and writes to DuckDB. Must always be running — Airflow executes dbt every 30 minutes via this container.

The pipeline scheduler. Runs 8 automated tasks every 30 minutes — seed mappings, dbt run (23 models), refresh PostgreSQL Gold, export to MinIO Silver/Gold, sync Metabase schema, column alerts and completion log. PostgreSQL backend with LocalExecutor.

Survey findings, geographic coverage, gender distribution, satisfaction scores and data quality for all surveys submitted by institution researchers.

My submissions, data quality scores, respondent distribution by county and benchmarking against national averages.

Total beneficiaries reached, counties covered, research areas funded, geographic reach and data submission accountability.

Pipeline health, ingestion success rates, bronze file registry, researcher activity and full audit trail of all ingested data.

JupyterHub collaborative notebook environment. Institution → Department → Project structure. Supports Python, R and SQL. Connected to Bronze (read), Silver and Gold (read/write). Auto-created on survey submission.

Internal analysis dashboard combining Kenya Tertiary Education and Kenya Economic survey findings. Includes institution satisfaction rankings, economy ratings, geographic reach and demographic breakdowns.

Download original Parquet files as submitted by researchers. Available as Parquet or CSV.

Download cleaned and standardised survey datasets — one file per survey family. Available as Parquet or CSV.

Download aggregated KPI tables ready for analysis in R, Python or Excel. Includes satisfaction scores, geographic distribution, demographics and programme analysis.