Microcycler Python SDK

Python SDK for the Microcycler battery testing platform.

Status: Pre-alpha — API subject to change.

Installation

pip install microcycler

The package is a thin Python layer over a Rust core (core/ in the repository: the same code the Microcycler dashboard and the headless agent run), shipped as one prebuilt wheel per platform (Windows, macOS, Linux; Python 3.9+). pip install microcycler needs no compiler. Installing from a source checkout (pip install -e sdk) builds the core and needs a Rust toolchain (rustup.rs).

Two ways to run an experiment

• Standard: build a recipe from steps (CC, CV, rest, loop, EIS, …) and store it on the cycler. It runs on the cycler by itself, so the computer can be switched off (the Recipes section below).
• Dynamic: a Python script on a host computer drives the cycler directly over TCP/UDP, for a test that does not fit the building blocks. The computer must stay connected: if the script dies or the host goes quiet for 30 s, the cycler releases the output on its own. The cycler reports its state as DYNAMIC while a script holds it. Scripts are never uploaded to a cycler; one process runs per cycler:

microcycler-dynamic run my_test.py --cycler 192.168.1.50 --cycler 192.168.1.51 --cell-id NMC-A1-001 --cell-id NMC-A1-002 --run-id run-7

from microcycler.dynamic import context

async def main():
    ctx = context()                                   # the cycler this process drives, and the test's metadata
    async with ctx.cycler() as c, c.dynamic(voltage_upper=4.2, voltage_lower=3.0, current_upper=0.5) as out:
        await out.set_current(0.2)
        ...

C-rates and the Capacity Check

Wherever a step takes a current you can give amps or a C-rate: CC(current=0.25) is 250 mA, CC(current=C(1)) is 1 C (250 mA for a 250 mAh cell) and C(0.25) is a quarter of that. The cycler turns a C-rate into amps when the test starts, from the capacity you give it, so one recipe serves cells of any size; start_recipe refuses (before sending anything) a recipe that uses one without capacity_mah:

from microcycler.recipes import C, CC, CapacityCheck, Recipe

recipe = Recipe("Characterise", steps=[CapacityCheck(charge_current=C(0.25), discharge_current=C(0.25), cutoff_current=C(0.05)), CC(current=C(1), cutoff_voltage_high=4.2)])
await cycler.start_recipe("characterise.json", cell_id="NMC-A1-001", run_id="run-7", content=recipe,
                          capacity_mah=250, v_min=2.5, v_max=4.2)

CapacityCheck(charge_current=..., discharge_current=..., cutoff_current=...) charges to the top voltage, holds until the current tapers to the cutoff (default C/20), rests, and discharges to the bottom voltage; the discharge counter starts from zero, so what it delivers is the result. The capacity it delivers is saved on the cycler's SD card as one small artifact file (<result file>_CAP_sNN.csv), so the result is kept whether or not a computer is connected; the dashboard adds it to the cell's record when it next sees the cycler. Leave voltage_max / voltage_min out to use the cell's own window (v_min / v_max).

Recipes on the cycler

A cycler stores recipes on its SD card (recipes/<name>.json); upload one once and any client can start it by name:

name = await cycler.upload_recipe(Recipe("NMC formation", steps=[...]))     # -> "NMC_formation.json"
print(await cycler.recipe_files())
await cycler.start_recipe(name, cell_id="NMC-A1-001", run_id="run-7", test_type="formation")

cell_id, run_id and test_type (1-24 characters from A-Za-z0-9._-) name the data files the test writes under data/ on the card. The dashboard and the phone start recipes the same way.

Uplink agent

Set up a headless computer (for example a Linux box in the lab) in two steps, with no browser and nothing to copy:

pip install microcycler
microcycler-agent enroll --install-service     # prints a code like ABCD-EFGH

An admin enters the code in the dashboard (Settings → Organization → Uplink → Add a computer) after checking the fingerprint, and the computer then uploads this network's licensed cyclers to your organization as a background service. Whether uplink is persistent is explicit: enroll and import ask "Keep uplink running in the background and start it at boot?" (or pass --service / --no-service), microcycler-agent run says on its first line whether it is running in the foreground (stops when the terminal closes) or as a service, every status line carries mode and persistent, and microcycler-agent status shows the service state. Only one agent runs per computer: starting a second one (another terminal, the service, or the dashboard) is refused with who already has it. microcycler-agent run runs it in the foreground instead; import <credential> is the offline alternative for a credential file made in the dashboard. Local telemetry needs UDP 5001/5002 open on a host firewall (sudo ufw allow in proto udp from 169.254.0.0/16 to any port 5001,5002, adjusting the source range to your cyclers' subnet).

Quickstart

Connect and read telemetry

import asyncio
from microcycler import Cycler

async def main():
    async with Cycler("192.168.1.100") as cycler:
        print(await cycler.identify())
        t = await cycler.get_telemetry()
        print(f"State:    {t.state}")
        print(f"Voltage:  {t.voltage:.3f} V")
        print(f"Current:  {t.current:.3f} A")
        print(f"Temp:     {t.temperature:.1f} °C")
        print(f"Capacity: {t.capacity_mah:.1f} mAh")

asyncio.run(main())

Discover cyclers on the network

import asyncio
from microcycler import Cycler, discover

async def main():
    # Listens for the discovery beacon on UDP 5002. The beacon is periodic
    # and unsolicited, so this is a collection window, not a deadline —
    # allow at least one beacon period (10 s by default).
    for dev in await discover():
        print(dev)                      # HADES-042 @ 169.254.10.5:5000 (standby)
        async with Cycler(dev.host, port=dev.port) as cycler:
            print(await cycler.get_life())

asyncio.run(main())

Stream live telemetry via UDP

from microcycler.streaming import CyclerStream

async def monitor():
    async with CyclerStream() as stream:
        async for telemetry in stream:
            print(telemetry.voltage, telemetry.current)

Build and run a recipe

from microcycler import Cycler
from microcycler.recipes import Recipe, CC, CV, Rest, Loop

recipe = (
    Recipe(name="Standard Formation")
    .add(CC(current=0.1, cutoff_voltage_high=4.2))
    .add(CV(voltage=4.2, limit_current_high=0.1, cutoff_current_low=0.02))
    .add(Rest(duration=1800))
    .add(CC(current=-0.2, cutoff_voltage_low=2.8))
    .add(Rest(duration=1800))
    .add(Loop(cycles=3))
)

async def run():
    async with Cycler("192.168.1.100") as cycler:
        await cycler.load_recipe(recipe)
        await cycler.start()

Recipe steps

License

MIT

Lifetime wear counters

Every board keeps an odometer in its EEPROM. Three independent axes, because a unit can be worn out along any one of them while the others look fine: mechanical (relay actuations), calendar (powered time) and electrical (charge through the output).

from microcycler import Cycler

async with Cycler("192.168.1.200") as cycler:
    life = await cycler.get_life()

    print(f"boots            {life.boot_count}")
    print(f"powered          {life.powered_hours:,.0f} h")
    print(f"throughput       {life.throughput_ah:,.1f} Ah")
    print(f"MAIN relay       {life.relay_main:,} closings")
    print(f"FAN relay        {life.relay_fan:,} closings")

    if not life.durable:
        print("WARNING: counters are not reaching the EEPROM")

These belong to the physical board, not to your database, so they survive a reinstall, an MCU swap, or the unit moving to another lab. There is deliberately no way to reset them — an odometer that can be wound back is not evidence of anything.

life.durable (i.e. write_errors == 0) is the field worth alerting on. A unit with a dead EEPROM still counts perfectly well in RAM and reports entirely plausible numbers; nothing else distinguishes a kept odometer from one that is lost at every power cut.

Counters are flushed at most once every few minutes to spare the part, so up to that much accrued wear is in RAM only. Before deliberately powering a unit down:

await cycler.flush_life()