# Vector databases Vector databases store high-dimensional vectors and search them by similarity. SnoutData supports **Pinecone**: connect with an API key, see your indexes in the explorer, and query them with SQL. ## Connect to Pinecone 1. Open the **connections sidebar** and choose **New connection**. 2. Pick the **Pinecone** driver. 3. Enter an **environment** label and your **API key**. There is no host or port to configure; SnoutData discovers your index hosts from Pinecone. 4. Click **Test**, then **Save**. Your API key is stored encrypted by your operating system's keychain, the same as any other connection secret. See [Security](https://docs.snoutdata.com/connections/security). ## Browse indexes Expand the connection to list your project's **indexes**, shown like tables. Each index exposes columns for the record `id`, the similarity `score`, the stored `values` (the vector), and the metadata fields found on its records. ## Query with SQL Query an index with SQL. The `WHERE` clause becomes a Pinecone **metadata filter** and `LIMIT` sets how many matches to return (`topK`): ```sql SELECT id, score, title FROM articles WHERE category = 'news' LIMIT 10; ``` Supported `WHERE` operators include `=`, `!=`, comparisons, `IN` / `NOT IN`, and `IS [NOT] NULL` (which maps to a field-exists check). `JOIN`, `GROUP BY`, `ORDER BY`, `OFFSET`, and `LIKE` are not supported on a vector query. ## On the roadmap Two vector features are not available yet: - **Semantic search from plain text** ("find records similar to this sentence"). This needs an embedding model to turn your text into a query vector, which is coming. - A **native vector-query editor** (the equivalent of the MongoDB pipeline editor). For now, query indexes with SQL. Because a vector query needs an actual query vector, there is no AI SQL-to-vector translation: SQL on a Pinecone connection always uses the offline compiler.