# Querying beyond SQL SnoutData works with three kinds of database: **relational** (SQL), **document** (MongoDB), and **vector** (Pinecone). A connection has one native execution form, and you choose how to author a query on top of it. There are three ways, and they all run as what the database actually executes. ## SQL, everywhere SQL is the common language. On a relational database it runs as SQL. On a document or vector database, SnoutData **transcribes** your SQL to the database's native form: - MongoDB: a SQL `SELECT` becomes an **aggregation pipeline**. - Pinecone: a SQL `SELECT` becomes a **vector query** (a metadata filter plus a result limit). The common shapes (selecting columns, `WHERE`, `ORDER BY`, `LIMIT`, `GROUP BY` with simple aggregates) are translated **exactly and offline** by a built-in compiler, so browsing and everyday queries cost nothing extra and never leave your machine. ### When SQL is too complex to translate offline If a query is more than the offline compiler can represent on a document database, SnoutData can ask the **AI assistant** to translate it. Because that uses your AI budget, SnoutData asks first; a plain **Run** never spends your budget silently. You can also opt in to always translate with AI from **Settings** if you prefer. (On vector databases there is no AI fallback, because a vector query needs an actual query vector that the model cannot invent.) ## Native When you want the full power of the database, write its native query directly. On a MongoDB connection the editor toolbar has a **SQL / Pipeline** switch: flip to **Pipeline** and write a real [aggregation pipeline](https://docs.snoutdata.com/databases/mongodb#write-a-native-pipeline) as a JSON array of stages, with schema-aware completion. The result lands in the same grid as a SQL query. ## Plain language Ask the [AI assistant](https://docs.snoutdata.com/ai-assistant/overview) in plain English. It is grounded in the connection's schema (collections and fields for MongoDB) and proposes a query you can run: SQL on a relational database, or a native aggregation pipeline on a document database. ## Where to go next - [MongoDB](https://docs.snoutdata.com/databases/mongodb): browse collections, query with SQL or a native pipeline, edit documents, and manage collections and indexes. - [Vector databases](https://docs.snoutdata.com/databases/vector): connect to Pinecone and query your indexes.