SuperAPI AI Directory

pp-tabelog

General & other

View repository

Updated 10-16 01:32

About this skill

Tabelog

Use tabelog-pp-cli from PATH, or ./tabelog-pp-cli in a built checkout. Build instructions: README.md. Public English sources need no account or API key.

Prerequisites: Install the CLI

This skill drives the tabelog-pp-cli binary. You must verify the CLI is installed before invoking any command from this skill. If it is missing, install it first:

After this CLI is merged into the public library, install it with the Printing Press installer:

npx -y @mvanhorn/printing-press-library install tabelog --cli-only

If the installer is unavailable, install directly with Go 1.26.6 or newer:

go install github.com/mvanhorn/printing-press-library/library/food-and-dining/tabelog/cmd/tabelog-pp-cli@latest

Go installs the binary into $GOBIN when set, otherwise $GOPATH/bin (default $HOME/go/bin). Add that directory to $PATH for the agent/runtime that will invoke this skill, or invoke the installed binary by its full path.

Verify: tabelog-pp-cli --version

If --version reports "command not found" after install, the runtime cannot see the binary directory on $PATH. Do not proceed with skill commands until verification succeeds.

Workflow

  1. Resolve a location with areas and a category with cuisines. Use returned typed choices when names are ambiguous.
  2. Use find for a small source-ranked candidate set; inspect selected canonical URLs or cached IDs with show.
  3. Save fetched candidates with lists add. Maintain notes and membership with lists show, lists note, and lists remove.
  4. Compare saved facts offline. Audit missing or old evidence before refreshing selected IDs.

Completion means structured results and their provenance support the answer. Use command help for current flags instead of loading the full command catalog.

Compact discovery

tabelog-pp-cli find --area tokyo --cuisine bar --meal dinner --budget-max 5000 --limit 5 --agent --select items.id,items.name,items.rating,items.review_count,items.dinner_budget,items.url

Read items for domain results and meta for effective criteria, source, age and coverage. --agent retains meaningful facts in compact JSON. Projection narrows fields while retaining essential metadata. A source next-page link is not a cursor immediately after the last displayed item.

Find summaries share identical fetched_at and source_surface values in meta; a row value overrides that default. meta.budget_source supplies an omitted meal-budget source; an explicit budget source overrides it. Explicit projection reads the full records.

Unique Capabilities

Command Use it for
lists add Persist fetched candidates and personal notes
lists compare Compare saved facts offline
lists refresh Fetch current details and report changes
lists alternatives Match saved backups against explicit constraints
lists audit Find unfetched, unknown or old evidence

For saved-candidate planning, read docs/saved-lists.md.

Accuracy and boundaries

  • Preserve ratings, review counts, separate meal budgets, links and retrieval times. Tabelog scores are weighted source signals, not universal quality verdicts.
  • Budget brackets are source estimates. Station meters are relative to the named station. Reservation policies do not establish current availability.
  • Distinguish source-unknown facts from details not fetched. Report bounded coverage, stale data and partial refresh failures. --data-source local makes zero source requests; a cache miss does not mean no restaurants exist.
  • On failure, inspect stdout for usable partial results and stderr for diagnostics. Domain errors are JSON under --agent; global flag-parser errors can be text.
  • Booking, payment/account changes, live availability/open-now conclusions, walking routes and bulk review/image harvesting are outside this CLI's scope.