v1.2.0 — data honesty + Smithery quality

npm: npm install -g tabedata-mcp · Smithery: listing next

A data-honesty + quality release. No nutrition value changed; what changed is how the data is described, labeled, and surfaced.

Data honesty

• The dataset is now described accurately as AI-compiled (an assistant reading cited sources), not individually human-verified. Renamed source.verified_at/by → compiled_at/by across all 460 records and the type. Removed the "no third-party / pure-curated / three-gates / verified-before-merge" overclaims from every doc.
• Reclassified ministop-fried-chicken-piece official_label → estimated (estimate by analogy; Ministop does not publish this SKU). Removed two malformed notes on 7eleven-onigiri-tuna-mayo. creatine-monohydrate-5g confidence high → medium.
• verify-data: a dead source.url is now a warning, not a build-blocking error.

Smithery quality

• READONLY annotations on all 6 tools, .meta({ title }) on all 24 parameters, server instructions, the plan_my_day prompt, and a smithery.yaml config schema with required: [].
• Removed stale pre-1.0 version references from descriptions, runtime output, and comments.

Release

• GitHub Actions CI (Node 20 + 22): typecheck, unit tests, build, data-shape validation.
• Project icon, README CI + downloads badges.

Verified: typecheck, 49/49 unit tests, integration against the built dist, verify-data 460/0/0, and a live MCP protocol probe (6 tools annotated, 24/24 params titled, prompt + instructions present). CI green on Node 20 and 22.

🤖 Generated with Claude Code