Installation

Fang is pure Python 3.11+ with no required dependencies.

pip install copperhead-fang

The distribution is copperhead-fang because the bare name on PyPI belongs to an unrelated package. The import name is still fang.

Two optional extras exist. When either is absent the toolchain says so rather than working around it:

pip install "copperhead-fang[analysis]"   # NetworkX, for graph analysis queries

ngspice is an external simulator reached across a process boundary. Install it with your package manager and fang will find it on PATH.

Your first board

divider.py

from fang.lang import System, V, kOhm, uF
from fang.parts import Capacitor, Resistor


class Divider(System):
    top = Resistor(resistance=10 * kOhm, package="R_0603_1608Metric")
    bottom = Resistor(resistance=4.7 * kOhm, package="R_0603_1608Metric")
    filter_cap = Capacitor(capacitance=100 * uF, voltage_rating=16 * V)

    def architecture(self):
        self.top.p2 >> self.bottom.p1
        self.bottom.p1 >> self.filter_cap.p1
        self.bottom.p2 >> self.filter_cap.p2

fang build   divider.py            # elaborate, pass the gate, persist, run the plan
fang netlist divider.py            # components and nets
fang export  divider.py -o out.net # a KiCad netlist

fang build writes a .copperhead/ workspace beside your sources. It holds the design as a canonical record stream, a cache that can be rebuilt and a manifest naming the schema version and the snapshot hash. Deleting the cache loses no engineering fact.