BlueBear Battery Research

A focused research companion to Battery Platform: current-limit models, condition/protocol matching, and an independent benchmark. This is experimental screening software, not a validated prediction of cell safety or lifetime. The 40 micrometer Li|Li6PS5Cl|Li target has not been experimentally validated.

Run locally

Python 3.11 or newer:

python -m venv .venv
# Activate .venv using the command appropriate for your shell.
python -m pip install -e ".[test]"
python -m pytest -q
python -B scripts/check_external_stripping_reference.py

The diagnostic reports four article-summary cases, three component predictions, one pressure-range abstention, and zero eligible independent accuracy cases. A null MAE means no eligible score, not zero prediction error. Temperatures in the sensitivity sweep are assumptions, not recorded specimen temperatures.

Research entry points

• services/shared/critical_current_density_policy.py: published current-limit observations, stripping equations, parameter intervals and applicability limits.
• services/shared/project_measured_limits.py: pure in-memory condition matching. collect_measured_current_limits accepts records supplied by the caller; there is no database adapter in this distribution.
• services/shared/schemas/measurement_protocol.py: nominal duration/capacity, explicit rest and stopping criteria. Actual elapsed time is a separate fact.
• scripts/current_limit_benchmark.py: exact errors and one-sided bounds, study-balanced aggregation, development-study exclusions and abstentions.

Equal declared protocols do not establish equivalence of missing preparation or hardware details. Legacy matching when neither side declares a protocol is compatibility behavior, not verified protocol equivalence. Matching tolerances are implementation choices requiring scientific review, not newly validated transfer laws. Fixed-current survival is not an exact critical-current failure.

Scope and provenance

This package is generated from the same implementation used by Battery Platform; there is no separately maintained copy of the science code. SOURCE_MANIFEST.json records SHA-256 hashes and source paths. No private Git history is included. Commercial optimization, experiment prioritization, decision workflows, the web application, database records, correspondence and lab quotes are excluded. Shared design DTOs are included to preserve compatibility; they do not include ranking or workflow services. Do not treat this as the full numerical solver.

The Zhao diagnostic cites DOI 10.1038/s41467-025-59567-8. It contains summarized observations, not a raw workbook or licensed data grant. Its existing reuse exclusion remains active. Source references and model-development anchors are identified in SOURCES.md, the model comments and diagnostic. No new data-use permission is implied by the code release. Report issues with a minimal synthetic example; do not submit private laboratory records or credentials.

Release status

This package is prepared locally for review. If LICENSE is absent, a software license has not yet been selected for this distribution. Before public release, include the selected LICENSE and review RELEASE_NOTES.md. The preparation tool does not publish repositories.