Publish reproducible chart-to-CSV fixtures before restoring accuracy claims

[kiku-jw]

Goal

Create a reproducible proof path for Chart2CSV extraction quality before publishing accuracy, speed, or comparison claims.

Required artifact

For each supported chart category (line, bar, scatter):

• source image with provenance;
• expected ground-truth CSV;
• actual extracted CSV;
• model/path name and exact version;
• run date and command or request;
• comparison method and mismatch report;
• clear distinction between browser Mistral extraction and local CV/Tesseract extraction.

Acceptance

• fixtures and expected outputs are checked into the repository;
• one repeatable command or test produces the actual outputs;
• results are dated and do not generalize beyond the tested fixture set;
• public README/site links to the artifact instead of claiming an unsupported universal accuracy percentage;
• any performance measurement records hardware, network/model latency boundary, and sample count.

The current site baseline intentionally labels the project Alpha and removes unsupported 90%+, fastest, competitor, and pricing claims until this evidence exists.

Tracked from kiku-jw/kikuai-project-map#13.