About the project
A card-payments company with POS terminals across the country needs to know where each terminal is, which merchant holds it, which SIM card is in it and what has happened to it since it arrived. Tech5A IMS follows every terminal from the head-office store to a region, to a merchant, to the repair desk and back, together with the SIM cards and spare parts that go with it.
Head office imports terminals from Excel in batches, verifies them and transfers them to a region in bulk. The region accepts or rejects each transfer, deploys accepted terminals to merchants with the merchant and terminal IDs, the SIM card, the acquirer, the store contact and the field specialist, and picks them up again with a reason. A picked-up terminal goes to the repair desk, which records the reported and diagnosed issues and the parts used, then marks it repaired and back with its region, or dismantled. SIM cards and spare parts move through their own head-office, regional and in-hand screens.
Regional managers see and change only their own regions' records, and the server refuses the rest whatever a screen sends. Roles grant view, create, edit, delete and approve on each of 28 screens. Every change goes on an audit trail, and the reports read from it: POS audit, user audit, terminal health, repair status, POS and terminal status, acquiring and the State Bank of Pakistan report, plus Dynamic Reports that pick any columns and filters. Every list filters, sorts and exports to Excel.
Optional AI helps the repair desk. On the Repair form, Suggest reads the pickup reason and the technician's notes and proposes one of ten fault categories with a one-line reason; the technician uses it or not. On a terminal's page, AI drafts a short history from the audit trail: its transfers, deployments, pickups and repairs, with their dates. Serial numbers, IMEIs, SIM and phone numbers, addresses and people's names are never sent, merchants go as M1 and M2, and a check refuses a draft that invents figures, money or verdicts. It runs on a self-hosted Ollama model by default or on one of sixteen providers, with a backup, daily limits and a budget, and drafts are made in the background so no page waits for them.
An activity log records every page opened, every change and download, refused requests and each AI call with what was sent and answered, keys masked, and an AI usage report shows the calls, tokens and cost by feature and by model. Staff sign in with Active Directory or with passwords the app stores hashed. A first-run setup wizard stores the database and directory settings encrypted, and App Settings changes them without a restart.
It is a React 19 and react-admin single-page app over an Express 5 API on PostgreSQL, with light and dark modes and fourteen colour palettes. It runs in Docker behind Caddy or nginx, or as a service on Ubuntu, and comes with user, technical and deployment manuals and 300 automated tests.
