# Load parquet into instant databases and query it

Source: https://www.hotdata.dev/use-cases/dataset-uploads
Site index: https://www.hotdata.dev/llms.txt

Parquet dumps and other one-offs usually need to live next to warehouse tables. Load them into an instant database, inspect the shape, then join them with everything else without standing up another pipeline.

## How it works

### Step 1: Upload from a file

**Claude**

```
Create an instant database with catalog q2, declare a targets table, and load targets.parquet into it. What do I type in SQL to read it?
```

**CLI**

```bash
hotdata databases create --catalog q2 --table targets
hotdata databases load --catalog q2 --table targets --file ./targets.parquet
```

**Python SDK**

```python
import hotdata

db_api = hotdata.DatabasesApi(api_client)
db = db_api.create_database(
    hotdata.CreateDatabaseRequest(
        name="Q2 targets",
        schemas=[hotdata.DatabaseDefaultSchemaDecl(
            name="public",
            tables=[hotdata.DatabaseDefaultTableDecl(name="targets")]
        )]
    )
)
uploads = hotdata.UploadsApi(api_client)
with open("targets.parquet", "rb") as f:
    up = uploads.upload_file(f.read())
conn_api = hotdata.ConnectionsApi(api_client)
conn_api.load_managed_table(
    db.default_connection_id, "public", "targets",
    hotdata.LoadManagedTableRequest(mode="replace", upload_id=up.upload_id),
)
# Query as: SELECT * FROM default.public.targets
```

### Step 2: Load more files into the same database

**Claude**

```
Load accounts.parquet as an accounts table into the same q2 database.
```

**CLI**

```bash
hotdata databases load --catalog q2 --table accounts --file ./accounts.parquet
```

**Python SDK**

```python
import hotdata

uploads = hotdata.UploadsApi(api_client)
with open("accounts.parquet", "rb") as f:
    up = uploads.upload_file(f.read())
conn_api = hotdata.ConnectionsApi(api_client)
conn_api.load_managed_table(
    "<db.default_connection_id>", "public", "accounts",
    hotdata.LoadManagedTableRequest(mode="replace", upload_id=up.upload_id),
)
```

### Step 3: Inspect before you join

**Claude**

```
Show my instant databases and list the tables inside q2.
```

**CLI**

```bash
hotdata databases list
hotdata databases tables list
```

**Python SDK**

```python
import hotdata

db_api = hotdata.DatabasesApi(api_client)
db_api.list_databases()
db_api.get_database("<database_id>")
```

### Step 4: Query like any other table

**Claude**

```
Match my Q2 targets to sales accounts on account id. Give me account_id and plan for up to 50 pairs.
```

**CLI**

```bash
hotdata query "SELECT t.account_id, s.plan FROM q2.public.targets t JOIN sales.public.accounts s ON s.id = t.account_id LIMIT 50"
```

**Python SDK**

```python
import hotdata

query_api = hotdata.QueryApi(api_client)
query_api.query(
    hotdata.QueryRequest(
        sql=(
            "SELECT t.account_id, s.plan FROM default.public.targets t "
            "JOIN sales.public.accounts s ON s.id = t.account_id LIMIT 50"
        ),
    ),
)
```

## Who uses this

- Ops bringing in recurring CSV extracts without touching warehouse DDL.
- Analytics freezing cohort definitions as stable instant-database tables.
- Automation that reads uploaded tables by name instead of parsing attachments ad hoc.