H1 Mirror loading the Python engine…
India's chemists already log every Schedule H1 antibiotic sale. H1 Mirror reads those lines and sends each registered prescriber a private monthly mirror: their Access / Watch / Reserve mix against same-specialty peers. This page runs the same Python engine as the repository (engine/domain + engine/application) inside your browser via Pyodide. Nothing you type leaves this page.
1 · Classify lines as written on an Indian bill or H1 register
One line per row: brand or generic name, strength and form. Brand → molecule (a table of 59,341 Indian antibiotic brands built from a public Indian medicine catalogue, plus field-study brands) → WHO AWaRe 2023 group, plus WHO's list of not-recommended fixed-dose combinations. A fixed rule table decides; no AI in this step. On that catalogue the rule table classifies 98.7% of 62,351 listed systemic antibiotic products (listed products, not sales; 23.9% are WHO-not-recommended combinations) and agrees with an independent blind reader on 200/200 randomly drawn products (engine/run_india_catalogue.py, audit/).
2 · One month of lines → private letters + district view
Load the synthetic sample month (12 pharmacies: 10 export complete bills, 2 only photograph their handwritten H1 register; 30 registered prescribers; unregistered providers; a few misread registration numbers) or your own CSV with columns date, pharmacy, prescriber_reg, specialty, drug_line, qty[, source]. What the number means: Watch + Reserve share = Watch + Reserve lines ÷ all WHO-classified antibiotic lines of one prescriber, counted only from complete bill exports (the H1 register alone holds mostly Watch drugs, so photo rows feed the district aggregate only). A letter needs ≥20 such lines and a registration number found on the council list; peers = same specialty. The district sees aggregates only when ≥10 pharmacies contribute.
3 · The chemist's side: a printable register, not extra work
Each chemist sees only their own month: every antibiotic line from bills and confirmed register rows, with its WHO group, ready to print for the drug inspector. Patient details stay on the chemist's own copy. No other pharmacy's data, and the district never gets a per-pharmacy report.
Code, tests and data sources: github.com/Ryugi62/h1-mirror · Idea submitted to Hack2Heal 2.0 (Global Healthcare Innovation Hackathon). The sample month is synthetic: no real pharmacy, prescriber or patient. WHO AWaRe classification 2023 (CC BY-NC-SA 3.0 IGO) · brand tables from Gautham et al., Antibiotics 2022;11(4):523 and the Indian Medicine Dataset (junioralive, GitHub, MIT licence).