Harden Solavin for research-grade I-V analysis

[dev-juy]

What changed

• Replaced fragile I–V calculations with a validation-aware analysis core using exact piecewise-linear maximum-power-point evaluation.
• Added explicit quality states for invalid, provisional, censored, and ambiguous sweeps; derived values are withheld when unsupported.
• Corrected sparse-channel voltage alignment, duplicate-voltage handling, signed-current behavior, SI scaling, and comparison logic.
• Removed chart smoothing that could invent curve shape and removed fabricated instrument/live/calibration claims.
• Hardened CSV/XLSX import and export, including formula-injection protection, size/row limits, provenance sheets, and efficiency inputs.
• Replaced the vulnerable XLSX dependency and upgraded the build/test stack.
• Added deterministic, status-aware research assistance that does not rank invalid or provisional results.
• Replaced the broken CI filename with a real GitHub Actions workflow and added citation and validation documentation.

Why

The prior implementation could silently convert missing measurements to zero, align channels by row rather than voltage, display an unbracketed open-circuit voltage as exact, calculate downstream metrics from unsupported values, and visually smooth data into shapes that were never measured. Those behaviors are unacceptable for research visualization.

Validation

• 53/53 automated tests pass.
• 300 randomized analytic linear-oracle trials.
• Six independent continuous single-diode/golden-search oracles.
• Scale-invariance, censored-sweep, ambiguous-crossing, malformed-input, sparse-alignment, export-provenance, and safety-limit tests.
• Coverage: 92.94% statements, 85.46% branches, 98.52% functions, 96.46% lines.
• npm audit --audit-level=high: 0 vulnerabilities.
• Standard and standalone production builds pass.
• CI installs Chromium and runs browser-level CSV/XLSX/chart smoke tests.

Research scope and limitations

This change makes the system fail transparently outside its validated scope; it does not claim IEC certification or universal correctness. Series and shunt resistance remain local diagnostic estimates, uncertainty propagation is not implemented, hysteresis requires separate forward/reverse sweeps, and instrument calibration remains the researcher’s responsibility. See docs/VALIDATION.md for the full validation register.

[netlify]

✅ Deploy Preview for solavin ready!

Name	Link
🔨 Latest commit	3afe2b5
🔍 Latest deploy log	https://app.netlify.com/projects/solavin/deploys/6a6b87afa8c2d90009f7aa83
😎 Deploy Preview	https://deploy-preview-2--solavin.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
🤖 Make changes	Run an agent on this branch

---

To edit notification comments on pull requests, go to your Netlify project configuration.