Data Manager — Sign in

Upload or paste CSV with a header row. Recognised columns: bemis (required), enrollment, level (P/M/H/S), gender (B/G/C), name. Only cells that differ become changes, and only for schools inside your permitted area — you review the diff before anything is saved.

Accounts & permissions. The super admin creates accounts and assigns each officer an area (divisions and/or districts) plus permissions. Officers only see and change schools inside their area; every saved change is stamped with their username. Use Export accounts to carry the same officer list to other computers.
Changes. Changes save on this device and apply instantly to every module. Export changes downloads sed-changes.json — a colleague merges it with Import changes. To publish permanently, the focal person runs the dataset build with the change file; the official schools-data.js is regenerated for everyone.
Honesty note. With no server, this is device-level access control — it organises and deters, it is not bank-grade security. The official data files remain read-only regardless of who signs in.

A register refresh is a whole extract, not a correction. The file must follow the sample format — columns bemis, name, district (required) and tehsil, gender, level, enrollment, lat, lng. Every row is validated before anything applies: unknown districts and missing BEMIS codes are rejected and listed, coordinates outside Balochistan are dropped, and a coordinate already held is never erased by a row that carries none. The official files are not touched — the import applies on top, can be removed, and becomes permanent only through the deploy file.

Describe the data first — the description is what every module will show — then upload a CSV of district,value or bemis,value rows. Once integrated it appears as an indicator in the Intelligence maps, the report’s choropleth panels and blind-spot audit, and the Assurance collection ranking. A district with no value reads as absent, not zero.

The PDF is read on this device — nothing is uploaded anywhere. The reader finds district names (through the same crosswalk the rest of the system uses, so “Hub” and “Usta Muhammad” resolve correctly) and proposes the number standing next to each; a person checks every value against the line it came from before anything integrates. Scanned-image PDFs carry no text and cannot be extracted — they can still be stored as reference documents.

How the deploy file works: it replaces js/import-overlay.js; then run python bump.py. Every device then carries the import without touching browser storage. What propagates live on import: the map, school lists and search, analytics headline figures (recomputed, not patched), the Assurance GPS worklist, report register sections, and — for datasets — the Intelligence indicators, report choropleths, blind spots and Assurance ranking. What waits for a Python rebuild: the 5-Year Plan baselines, integrity exposure, and the census comparison — they are build artifacts, and pretending an import moves them would be false.

Every page exposes window.SedAPI — a stable, versioned surface returning plain named objects, never the internal arrays. Any future tool, script or system should read through it; the JSON exports below carry the same shapes with provenance attached, so what leaves this system stays checkable.