Dev Tools
Snowflake web data enrichment
A Snowflake-native data app that uses `NIMBLE_AGENT_RUN` to run one `google_maps_search` Web Search Agent per warehouse row. Start with a table of search queries like "coffee shops in Williamsburg Brooklyn" or "gyms in Austin Texas," then generate a structured business universe with names, categories, addresses, ratings, review counts, phone numbers, websites, coordinates, and Google Maps URLs — all from SQL.
Quick Start
Inputs
- Search queries A list of location-based search strings inserted into the query staging table (e.g., “coffee shops in Williamsburg Brooklyn”, “gyms in Austin Texas”).
Outputs
- One Snowflake table with structured business results per search query
- Source query lineage per row
- Business identity: name, category, and address
- Reputation signals: rating, review count, and price level
- Contact and location: phone, website, coordinates, and Maps URL
- Operational fields: status, sponsored flag, and enrichment timestamp
- Full raw entity stored as raw_entity for future extraction
How it works
A 3-phase pipeline. Read the blog here for a deeper explanation.
- Set up Snowflake Run the setup script once asit creates the role, warehouse, database, secret, network rule, and external access integration Snowflake needs to call Nimble's API endpoint.
- Install NIMBLE_AGENT_RUN Run the UDTF script to install the Python table function. it takes an agent name and input payload, calls Nimble synchronously, and returns a Snowflake VARIANT.
- Load search queries Insert your location search strings into the query staging table (e.g., "coffee shops in Williamsburg Brooklyn").
- Run the enrichment A single SELECT calls NIMBLE_AGENT_RUN with google_maps_search once per row — Snowflake handles parallelism automatically.
- Flatten and write The VARIANT response is flattened into typed columns and written to LOCAL_BUSINESSES — no Python pipeline, no external orchestrator.
Stack
Nimble primitives plus the full runtime stack.
Nimble APIs
What it does
- NIMBLE_AGENT_RUN Snowflake UDTF that runs one Nimble Web Search Agent per warehouse row and returns structured fields as Snowflake VARIANT data.
- google_maps_search Discovers local businesses from live Google Maps search queries. Returns business entities under parsing:entities:SearchResult.
3rd Party Tools
Role
- Snowflake Stores the input query table, executes the enrichment SQL, flattens Google Maps results, and writes the LOCAL_BUSINESSES output table.
- Snowflake UDTF Runs NIMBLE_AGENT_RUN as a table function inside the SQL query.
- External Access Integration Allows Snowflake to securely call Nimble’s API endpoint from inside the warehouse.
- Snowflake Secret Stores the Nimble API key used by the UDTF.
- Snowflake Warehouse Provides compute for the enrichment job.
Reach out if you have any questions.
Talk to an Expert