dz-generics

Find the generic equivalents of an Algerian medicine from a brand name, a DCI, or a registration code — using the Ministry of Industry and Pharmaceutical Production's Nomenclature nationale des produits pharmaceutiques à usage de la médecine humaine.

An agent skill, a CLI, and a committed SQLite index of the registry.

$ python skills/dz-generics/scripts/lookup.py --name "DOLIPRANE"
Nomenclature nationale -- version Août 2026

Anchor: DOLIPRANE -- PARACETAMOL
  form         : COMRPIME
  dosage       : 1000MG  [key 1000MG]
  status       : active
  origin       : made in Algeria (ALGERIE)
  type         : GE (generic-equivalent)

EQUIVALENTS -- same DCI, same form (COMPRIME), same dose (1000MG):

   1. ANTALGAN -- 1000MG -- GE (generic-equivalent)
      made in Algeria -- SARL ALPHACARE
   2. DOLI-BIEN -- 1000MG -- GE (generic-equivalent)
      made in Algeria -- PHARMIDAL NS
   3. DOLIPRANE -- 1000MG -- GE (generic-equivalent)
      made in Algeria -- PROPHARMAL
   ...

What it does and does not do

It reports what the registry lists: which products share an active substance, form and dosage, which laboratory holds each one, whether it is made in Algeria or imported, and whether it is active, not renewed, or withdrawn.

It does not assess bioequivalence, excipients, stability, or clinical suitability. Equivalence here means same DCI, same form, same dose — nothing more. Substitution is the pharmacist's or prescriber's decision. This is a lookup aid, not a source of prescribing advice.

Install

# any agent, via the skills.sh CLI
npx skills add adelpro/dz-generics -l                 # preview
npx skills add adelpro/dz-generics -a claude-code --copy -y

# or just clone it
git clone https://github.com/adelpro/dz-generics

The repo is also an Agent Plugins 1.0.0 package (plugin.json), so compatible clients can load skills/ directly.

Query time needs only Python 3 and the standard library. Nothing to install.

Use

python skills/dz-generics/scripts/lookup.py --name "DOLIPRANE"   # brand
python skills/dz-generics/scripts/lookup.py --dci "PARACETAMOL"  # substance
python skills/dz-generics/scripts/lookup.py --code "03 B 081"    # reg. code
python skills/dz-generics/scripts/lookup.py --name "RIFEX 120" --json

On Windows, set $env:PYTHONIOENCODING="utf-8" first so accented and Arabic output is not mangled by the console codepage.

Exit codes: 0 answered (found or ambiguous), 1 not found, 2 usage error.

How a result is grouped

Five sections, and the difference between them is the point:

section	meaning
Equivalents	same DCI, same form, same dose. The only genuinely equivalent class.
Other dosages	same DCI and form, different dose. Not equivalent.
Unknown dose	same DCI and form, dose not determinable. Not evidence of a difference.
Other forms	same DCI, different form. Not equivalent.
Off-market	withdrawn or not renewed at the anchor's form and dose. Not available.
Other products under this name	the name covers more than one medicine. Not equivalents.

When one name means two medicines

Some brands cover more than one product with different active ingredients, and the tool reports that rather than silently picking one:

$ python skills/dz-generics/scripts/lookup.py --name "NOBAC"
Anchor: NOBAC -- ALGINATE DE SODIUM/BICARBONATE DE SODIUM/CARBONATE DE CALCIUM

EQUIVALENTS -- same DCI, same form (COMPRIME_A_CROQUER), same dose (500MG/267MG):
   1. NOBAC -- 500MG/267MG -- GE -- made in Algeria -- BIOPHARM

Other products under this name (different active ingredient set -- NOT equivalent):
   NOBAC ADULTE GOUT FRAISE -- ALGINATE DE SODIUM/BICARBONATE DE SODIUM -- SUSPENSION BUVABLE
   ...

NOBAC is a chewable tablet (alginate/bicarbonate/calcium carbonate) and a suspension (alginate/bicarbonate) — confirmed against the manufacturer's own product page and the French ANSM/HAS monograph for the same class, where Gaviscon's tablet carries calcium carbonate and its suspension does not.

The opposite case also occurs: MANTIXA carries terbinafine (an antifungal cream) and molsidomine (a heart tablet) — nothing in common. The tool says so. That is a fact about the registry, and it is the opposite of a "probable equivalent" suggestion: guessing that two products under one name must be related is exactly how a lookup tool starts inventing medicine.

Every off-market row carries [withdrawn] or [not renewed] wherever it appears; a row with no marker is active.

Three things the output is careful about, because each is a way to be confidently wrong about a medicine:

An unknown dose is not a different dose. q.s and multi-ingredient dosages have no determinable strength, so they get their own section instead of being reported as a different strength.
Off-market is never presented as an option. A withdrawn product is a safety fact, not a formatting detail.
A near match is not a match. A fuzzy hit lists candidates and stops; it never picks one for you.

Data provenance

Source: Ministère de l'Industrie Pharmaceutique, Nomenclature nationale.
Current index: version Août 2026, 9,595 registration rows (5,425 active / 1,491 not renewed / 2,679 withdrawn), built from clean_NOMENCLATURE.VERSION.AOUT_.2026-.xlsx.
The ministry's spreadsheet is not redistributed here. data/source/ is gitignored. scripts/fetch_source.py downloads it in one request, or fetch it by hand from the page above. What ships is the derived index: a factual list of registrations, with attribution.
The index records a version label and build date, and the tool warns when it is more than 90 days old. Check the version before relying on an answer.

Refreshing the data

The ministry publishes every one to two months.

python skills/dz-generics/scripts/fetch_source.py        # newest release
python skills/dz-generics/scripts/build_index.py <the .xlsx>

build_index.py fails loudly rather than indexing garbage: it resolves columns by header name per sheet, cross-checks each sheet against the row count the sheet declares for itself, and refuses to write when the row count collapses. scripts/profile_source.py prints the distributions it uses, for when the ministry renames something.

Development

python -m pytest -q -W error

80 tests. The suite skips narrowly when the gitignored source workbook is absent, so a fresh clone runs green.

The normalization layer is the riskiest part of this project: it decides when two spreadsheet rows describe the same medicine, and a wrong answer there is invisible. Its rules are derived from the observed data, not assumed, and references/schema.md documents the source layout they are built on.

License

MIT. See LICENSE.

The MIT licence covers this project's code and derived index. It does not grant rights to the ministry's spreadsheet, which is not included.