place-collector-viewer

name: place-collector-viewer description: > Collect Naver place datasets and inspect them in a local web datagrid. Use when asked to gather places from Naver Maps, maximize coverage in dense areas with bounds tile split + pagination + placeId dedupe, export JSON, compare collection methods, or view/filter place lists in place-datagrid-view.html. Triggers include "플레이스 수집", "장소 데이터 추출", "네이버 지도 데이터 모으기", "bounds 타일 분할", "placeId 중복 제거", "json 조회", "웹 데이터그리드로 장소 보기".

Place Collector Viewer

Use Chrome DevTools MCP to collect Naver place data, save it as JSON, then auto-open and auto-import it in the local datagrid viewer.

Goal

1. Collect a high-coverage place pool.
2. Open the viewer and import the JSON automatically.
3. Browse/filter/sort the imported pool.

Scope

• Include: place collection, coverage comparison, JSON export, viewer auto-loading.
• Exclude: weather, date-course planning, budget itinerary generation.

Preconditions

• Use Chrome DevTools MCP.
• Keep /Users/jongyeol/.claude/skills/place-collector-viewer/assets/place-datagrid-view.html available.

Non-negotiable Rules

1. Use bounds tile split + pagination + placeId dedupe as the default collection method.
2. Do not claim full coverage from one wide bounds query alone.
3. Retry failed MCP/script calls up to 3 times.
4. Do not report places that are not in the exported JSON.
5. When a JSON file path exists and user does not opt out, auto-open viewer and auto-import JSON (do not ask user to upload manually).

Workflow

1) Verify MCP

• Call list_pages.
• If unavailable, follow references/troubleshooting.md and stop if recovery fails.

2) Lock collection parameters

• queryTexts: e.g. 강남역 음식점, 강남역 카페
• center: {x, y} for the target place/area
• radiusMeters: default 300 (increase for validation runs)
• gridSize: default 6 (6~8 for dense zones)
• display: default 70
• output path: ~/Downloads/place-collector-viewer/dp-{area}-{YYYY-MM-DD}.json

3) Build base pool (required)

• Read references/naver-map-guide.md.
• Run:

1. Step 0 to clear localStorage.__dp__
2. Step 1A (tile split GraphQL collection with pagination)

• Expand queryTexts and rerun Step 1A when needed.

4) Compare coverage methods (when requested)

• Method A: tile split + pagination + dedupe (Step 1A baseline)
• Method B: one wide bounds + pagination + dedupe
• Compare placeId sets and report overlap/missing counts.

5) Enrich data (default on)

• Run Step 2 (getVisitorReviewStats, batch 50)
• Run Step 3 (getPlaceDetail, batch 50)
• Run Step 4 (getFeeds, batch 50 with sequential fallback)

6) Export JSON

• Run Step 5 to export __dp__.
• Ensure output directory exists (mkdir -p ~/Downloads/place-collector-viewer).
• If MCP output is wrapped or broken, clean with scripts/extract-dp.py.
• Save final path as outputJsonPath.

7) Auto-open viewer + auto-import JSON (default)

Run this automatically when outputJsonPath exists.

1. Open viewer page:

• file:///Users/jongyeol/.claude/skills/place-collector-viewer/assets/place-datagrid-view.html

2. Find file input with snapshot (#fileInput).
3. Upload file by MCP upload_file using outputJsonPath.
4. Verify load state by reading #status text (데이터 X / Y개).
5. If total count is 0 or parse fails, retry upload once.

Status check script:

() => {
  const text = document.querySelector('#status')?.innerText || '';
  const m = text.match(/데이터\s*([\d,]+)\s*\/\s*([\d,]+)개/);
  const toNum = (v) => Number(String(v || '').replace(/,/g, ''));
  return {
    statusText: text,
    shown: m ? toNum(m[1]) : null,
    total: m ? toNum(m[2]) : null
  };
}

8) Report

• Include parameters (queryTexts, center, radiusMeters, gridSize, display)
• Include counts (totalUnique, retry/error summary, enrichment status)
• Include outputJsonPath
• Include viewer load result (shown/total)

Fast Paths

• If user already provides a JSON path, skip collection and run only Step 7 + Step 8.
• If user asks extraction only, skip Step 7 and return JSON path + stats.

Resources

• references/naver-map-guide.md: collection/enrichment/export scripts
• references/troubleshooting.md: MCP recovery steps
• scripts/extract-dp.py: JSON cleanup helper for MCP output
• assets/place-datagrid-view.html: local datagrid viewer