local-places

What This Skill Does

The local-places skill serves as a high-performance interface for the Google Places API, wrapped in a local proxy server. It enables the OpenClaw agent to perform context-aware location lookups and real-time place discovery. By leveraging a two-step resolution process—first grounding the user's natural language location input, then executing filtered geospatial searches—it ensures high accuracy for neighborhood-specific queries. The skill supports advanced parameter filtering, including minimum ratings, price levels, and business status (e.g., 'open now'), making it an essential tool for location-based assistance.

Installation

To integrate this skill, ensure you have the OpenClaw environment ready. Execute the installation via the CLI: clawhub install openclaw/openclaw/skills/local-places. Once installed, navigate to the base directory, configure your API key by running echo "GOOGLE_PLACES_API_KEY=your-key" > .env, and spin up the microservice using uv run --env-file .env uvicorn local_places.main:app. Ensure your environment has the required dependencies installed via uv pip install -e ".[dev]".

Use Cases

This skill is perfect for autonomous travel planning, recommendation engines, and logistics. It excels in scenarios where a user requires local discovery, such as identifying the nearest highly-rated coffee shops for remote work, finding open restaurants after midnight, or checking price tiers for dining out. Its ability to handle location bias makes it ideal for local concierge-style workflows.

Example Prompts

1. "Find me a coffee shop in Soho that is open right now and has at least a 4-star rating."
2. "I am looking for a restaurant in downtown Tokyo with a price level of 3, preferably Italian cuisine."
3. "List the top 5 gym locations near my current area in Brooklyn."

Tips & Limitations

To maximize the utility of local-places, always prefer specific type filtering (e.g., 'cafe' over 'food') to reduce noise. Note that the API is subject to standard Google Places constraints; therefore, filters.types must be singular for optimal query performance. If a search yields too many results, use the next_page_token to paginate effectively. Remember that location resolution is a separate step; if the agent struggles to find a spot, force a resolution call before searching. Be mindful of radius constraints to avoid empty result sets.