Version 0.4.1

Version 0.4.1#

Version 0.4.1 adds a driver-side power estimation subpackage, explicit controls over parallel execution, a packaged cookbook for AI coding agents, and restores compatibility with plotly 7.

Highlights#

  • New ccp.drivers subpackage with InductionMotor to estimate the shaft power delivered by the driver from field measurements and compare it with the thermodynamic shaft power computed by ccp.

  • Parallel execution can now be disabled or capped globally (ccp.config.PARALLEL, ccp.config.POOL_SIZE), and a script missing the if __name__ == "__main__" guard fails fast with an actionable error instead of hanging.

  • A ccp cookbook is shipped with the package as an agent skill and can be installed with the ccp-install-skill command.

  • import ccp works on plotly 7.

New features#

  • Drivers subpackage with induction motor power estimation (#155) — ccp.drivers estimates the shaft power delivered by the compressor driver so it can be compared with Point.power_shaft. InductionMotor implements the DOE/IEEE 112 field methods (input power, line current and slip, with voltage compensation and part-load efficiency curves) and supports VFD operation through supply_frequency and vfd_efficiency. The Driver.compare() helper returns a PowerComparison with driver power, power transmitted through coupling/gearbox, point shaft power and the absolute and relative deltas. Drivers are serializable to JSON and TOML. New electrical entries (voltage, current, power factor, frequency, efficiency) were added to the units checked by check_units.

  • Parallel execution controls and fail-fast pool startup (#158) — every multiprocessing pool created by ccp (Impeller.convert_from, Impeller.load_from_dict, Impeller.load_from_engauge_csv, Evaluation) now goes through ccp.parallel.create_pool():

    import ccp

ccp.config.PARALLEL = False  # run every ccp calculation serially
ccp.config.POOL_SIZE = 4  # or cap the number of worker processes

    The CCP_PARALLEL, CCP_POOL_SIZE and CCP_POOL_START_TIMEOUT environment variables override the globals, so parallelism can be tuned in containers and sandboxes without touching code. Worker startup is verified with a no-op task: if a full worker generation dies before it completes (the symptom of a missing if __name__ == "__main__" guard), ccp raises a RuntimeError pointing at the cause and the serial escape hatch instead of repopulating dying workers forever. Invalid settings raise a ValueError naming the offending variable.

  • ccp cookbook as an agent skill (#157) — nine self-contained recipe files plus a SKILL.md entry point, following the Agent Skills open standard, cover states and fluids, performance points, impellers, Engauge curve import, conversion to new suction conditions, ASME PTC 10 similarity, flow orifice metering, operational data evaluation and common gotchas. The recipes ship in the wheel, and the new ccp-install-skill console script installs them into the personal skills directory of each detected AI coding agent (Claude Code, GitHub Copilot, Cursor, Codex), with --agent, --project, --dest and --uninstall options. The installed SKILL.md is stamped with the ccp version so agents can flag a stale skill after an upgrade.

Bug fixes#

  • plotly 7 compatibility (#162) — plotly 7.0.0 removed the mapbox layout key and the scattermapbox trace type, which the ccp plotly template still registered, so import ccp raised ValueError: Invalid property specified ... 'scattermapbox'. The entries were dropped; the template loads on plotly 5.x through 7.

  • Module doctests (#156) — fixed the nine module doctests in processing.py, units.py, fo.py and impeller.py that failed when collected with --doctest-modules.

  • Docs dark mode (#159) — the landing-page card icons no longer render on white boxes in dark mode.

Infrastructure#

  • README badges updated: ruff replaces black, and a CI status badge was added (#154).