Brightway MCP validation and limitations

What the repository reliability checks cover, and what they do not establish.

Cross-engine reference case

The reliability suite uses the same two-process synthetic system in Brightway and openLCA. For 1 kg of reference product, the hand-derived total is 6.5 kg CO2e; the recorded Brightway result is 6.5 kg CO2e. The 2 kg scaling case returns 13.0 kg CO2e.

text

3 kg material × 2.0 kg CO2e/kg
+ 0.5 kg direct CO2e
= 6.5 kg CO2e

Additional checks

Deterministic repeated calculations.
Contribution totals reconciling with the total score.
Monte Carlo percentile ordering and basic sanity.
Negative technosphere exchange/sign behavior.
A network-gated characterization-factor spot check.

The recorded post-hardening suite reports 30 passing tests with one network test deselected by default.

What the tests do not establish

Correctness of every Brightway database or LCIA method.
Validity of an arbitrary foreground model.
Methodological compliance of a study.
Suitability of AI-generated model definitions.
Hosted uptime, security, or multi-user isolation.

Reproduce the checks

bash

python -m pytest -q
python -m pytest -m network -v

Note

The network-marked test can download external background data, so run it deliberately.