Query Yandex Search Results

This skill parses Yandex search engine result pages through Yandex Cloud Search API v2. It returns search results with position, title, URL, snippet, and domain for each entry. Both synchronous (one query at a time) and asynchronous (batch) modes are available. Results are cached in the cache/results/ directory.

Use this skill when you need to parse Yandex search results for specific queries with region targeting. The skill supports Russian regions including Moscow (213), Saint Petersburg (2), Kazan (43), and others listed in the regions table.

Do NOT use this if you don't have a Yandex Cloud service account configured. The skill requires a valid service account key file and folder ID in the configuration.

Users provide a search query (single text or file with multiple queries) and optional parameters including region ID, results per page (1-100), page number, search type (SEARCH_TYPE_RU/TR/COM/KK/BE/UZ), and family mode (FAMILY_MODE_NONE/MODERATE/STRICT). Region names can be resolved to IDs:

bash scripts/search_region.sh --name "Казань"

For single queries:

bash scripts/web_search_sync.sh \
  --query "купить дымоход" \
  --region-id 213

For batch processing:

bash scripts/web_search_async.sh \
  --file queries.txt \
  --region-id 213

Outputs are JSON arrays with position, title, URL, snippet (up to 300 characters), and domain. Raw XML responses and parsed JSON are saved to cache/results/<hash>.raw and cache/results/<hash>.json.

Requires Yandex Cloud service account with Search API v2 access. Configuration requires the Yandex Cloud folder ID in config.json and service account key file in config/service_account_key.json. IAM tokens are generated and cached automatically:

bash scripts/iam_token_get.sh

On macOS, OpenSSL installation via Homebrew may be required. The skill includes regions_tree.sh for viewing region IDs and search_region.sh for finding region IDs by name.

Users who need to parse Yandex search results with region-specific targeting. The synchronous mode handles single queries with immediate results, while asynchronous mode processes batches with automatic polling. Async operations can take minutes to hours; interrupted jobs can be resumed with the --resume flag.