> ## Documentation Index
> Fetch the complete documentation index at: https://docs.myquery.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# DuckDB

> DuckDB is an analytical database designed for lightning-fast queries on local files, widely used in data science and local analytics.

## **Security & Architecture Overview**

Before connecting your data, it is critical to understand how MyQuery interacts with your infrastructure:

* **Schema-Only Access:** MyQuery does **not** ingest or store your raw database records. We only read the metadata (table names, column types, and foreign key relationships) to build a semantic map for the AI.
* **Encrypted Storage:** All connection credentials and "maps" are encrypted at rest using industry-standard AES-256 encryption.
* **Whitelisting:** To allow MyQuery to reach your database, you must whitelist our outbound IP addresses in your firewall or cloud security groups (e.g., AWS Security Groups, Azure Firewall).

#### **Method A: Database File (.duckdb)**

Use this if you have an existing DuckDB database file.

1. Navigate to the **DuckDB** tab.
2. Click **Upload File** under the **Database File** tab.
3. Upload your

   ```
   .duckdb
   ```

   or

   ```
   .db
   ```

   file.
4. MyQuery will mount the file and allow you to select the internal tables for AI training.

#### **Method B: CSV or Parquet Files (Multi-File)**

This unique method allows you to create a virtual database from raw files.

1. Go to the **CSV or Parquet Files** tab.
2. Select **Upload Files** and choose one or more

   ```
   .csv
   ```

   or

   ```
   .parquet
   ```

   files.
3. **How it works:** MyQuery creates a temporary, in-memory DuckDB instance where **each file becomes a table**.
   * *Example:* If you upload

     ```
     customers.csv
     ```

     and

     ```
     orders.parquet
     ```

     , you will have a

     ```
     customers
     ```

     table and an

     ```
     orders
     ```

     table ready for natural language questioning.