enrich-place-details

name	enrich-place-details
description	Fetch enriched restaurant place details from the hosted Dineway API. Use when the user provides a restaurant name and city, optionally with a Google placeId, and Codex needs to create a shadow user, search Dineway places, select the correct match, re-bind auth to the final place id, save the full place details JSON to the current workspace, and return the saved file path.

Enrich Place Details

Use scripts/enrich_place_details.js to fetch restaurant details from the hosted Dineway API.

Required Inputs

The user must provide both:

• restaurant name
• city or city/country

If either is missing, ask for the missing value and stop the skill. Do not run the script with incomplete inputs.

placeId is optional. If the user provides a valid place id, pass it with --place-id and treat it as authoritative. If no valid place id is provided, the script uses a temporary bootstrap place id only to create the first shadow user, searches by restaurant name and city, then creates a new shadow user bound to the selected place id before fetching details.

Workflow

Run the script from the workspace where the place JSON should be saved. By default it creates places/${placeId}.json under the current directory and prints only a short summary.

node "$SKILL_ROOT/scripts/enrich_place_details.js" \
  --name "Peace Harmony" \
  --city "Sydney, Australia"

With a user-provided place id:

node "$SKILL_ROOT/scripts/enrich_place_details.js" \
  --name "Peace Harmony" \
  --city "Sydney, Australia" \
  --place-id "ChIJs5ydyTiuEmsR0fRSlU0C7k0"

Inspect candidates when matching is ambiguous:

node "$SKILL_ROOT/scripts/enrich_place_details.js" \
  --name "Peace Harmony" \
  --city "Sydney, Australia" \
  --include-candidates

Result

Read placeDetailsPath from stdout and use that file for downstream site generation. The saved file contains the full payload by default: query, authFlow, search, selectedPlace, and placeDetails. The script never prints access or refresh tokens.

Stdout is intentionally compact:

{
	"placeId": "ChIJs5ydyTiuEmsR0fRSlU0C7k0",
	"placeDetailsPath": "places/ChIJs5ydyTiuEmsR0fRSlU0C7k0.json",
	"savedPayload": "fullPayload",
	"selectedPlace": { "...": "selected place summary" },
	"search": { "...": "candidate count, selected candidate, match score, and match reason" }
}

Use --output-dir DIR only when a different destination is explicitly needed. Use --details-only only when the saved JSON should be the detail endpoint response instead of the full skill payload.

Matching Rules

• Prefer the user-provided place id when supplied.
• Otherwise score search candidates by Unicode-normalized name, location, food-related type, business status, rating, and review count.
• Name matching reads displayName.text, name, businessName, and title.
• Location matching reads formattedAddress, shortFormattedAddress, address, vicinity, plusCode, and addressComponents; city/country input is split into parts so Sydney, Australia can still match an address that only includes Sydney.
• Rating and review count are tie-breakers, capped so they do not override strong name/location evidence.
• Use id or placeId from the selected search result as the final Dineway detail lookup id.
• If the final id differs from the initial bootstrap id, or the initial id was only a bootstrap fallback, create a new shadow user for the final id before fetching details.
• When matching is ambiguous, rerun with --include-candidates and inspect matchScore and matchReason in each candidate summary.

Resources

• scripts/enrich_place_details.js: Node.js implementation of the shadow-user, search, rebind, refresh, and detail flow.
• references/api.md: Observed hosted API response fields and endpoint notes.