Prompt Optimizer v2.7.0

Summary

• Added a structured compare evaluation workflow so prompt comparison is easier to test, calibrate, and trust.
• Smoothed the compare experience across test panels, fallback logic, and workspace reactivity.
• Refreshed docs, marketing surfaces, and extension assets so the project is easier to discover and explain.

Highlights

• Introduced structured compare evaluation as a first-class workflow for prompt testing and analysis.
• Made compare UX feel lighter and more reliable across result panels and workspace state.
• Refreshed the public-facing surfaces with a standalone marketing site, updated demos, and localized extension store assets.

Product Updates

Web

• Added the structured compare workflow to test panels and simplified the compare evaluation experience.
• Kept workspace test state reactive across language switches and restore flows.
• Refined the About / resources area and removed compare hints when the current context does not support them.

Extension

• Added localized store screenshots and refreshed derived logo assets for store-facing materials.

Core/Infra

• Added a structured compare evaluation pipeline and calibration artifacts to support more stable comparison output.
• Refined evaluation rewrite behavior and aligned compare design docs with the new workflow.
• Improved docs build reliability while redesigning the onboarding and bilingual docs experience.

Fixes

• Hid misleading teacher-style compare hints when no reference context is available.
• Fixed compare-version fallback to use the conversation record ID more reliably.
• Cleaned up MkDocs build issues that were blocking parts of the docs deployment flow.

Breaking Changes / Upgrade Notes

• None. Existing data and workflows should continue to work without manual migration.

Developer Notes

• This release is the main handoff point for the structured compare architecture. If you maintain evaluation prompts or compare-related UI, use this version as the new baseline.
• The commit range also includes marketing-site and documentation work, but the dominant product change is the structured compare stack.