# Redis configurations

Available in v1 | Local development

This page describes how dbt models behave with the community-maintained `dbt-redis-adbc` adapter. The adapter's Redis ADBC driver presents Redis data as SQL relations named `schema.table`, in a single catalog named `redis`. For the general configuration syntax, refer to [materializations](../../docs/build/materializations.md) and [incremental models](../../docs/build/incremental-models.md). For installation and connection settings, refer to [Redis setup](../../docs/local/connect-data-platform/redis-setup.md).

## Materializations

Set `materialized` in a model's `config()` block or in `dbt_project.yml`. For the syntax, refer to [materializations](../../docs/build/materializations.md). The adapter supports these materializations:

| Materialization     | Support                                |
| ------------------- | -------------------------------------- |
| `table`             | SQL and Python models                  |
| `view`              | SQL models                             |
| `ephemeral`         | SQL models                             |
| `incremental`       | SQL models, using the strategies below |
| `materialized_view` | Not supported                          |

## Incremental models

Supported strategies are `append`, `delete+insert`, `merge`, and `microbatch`. With a `unique_key`, the default strategy is `delete+insert`. Without a `unique_key`, the default appends rows.

For example, use `merge` to update existing rows and insert new rows:

models/orders.sql

```sql
{{ config(
    materialized='incremental',
    incremental_strategy='merge',
    unique_key='order_id'
) }}

select *
from {{ ref('stg_orders') }}
{% if is_incremental() %}
where updated_at >= (select max(updated_at) from {{ this }})
{% endif %}
```

The adapter also supports [`on_schema_change`](../../docs/build/incremental-models.md#what-if-the-columns-of-my-incremental-model-change) and [`incremental_predicates`](../../docs/build/incremental-strategy.md#about-incremental_predicates). For microbatch models, configure `event_time`, `begin`, and `batch_size` using the standard [microbatch configuration](../../docs/build/incremental-microbatch.md). Each batch replaces its event-time window; batches can run in parallel.

## Model contracts and constraints

You can enforce model contracts. The driver enforces `not_null` and `check` constraints. It accepts `primary_key`, `unique`, and `foreign_key` constraints without enforcing them; dbt warns about those constraints.

Small integer literals have the `integer` type. In a contract, declare a column such as `select 1 as id` as `integer`, or cast the expression to `bigint` in the model.

## Persisted documentation

Enable `persist_docs` to store relation and column descriptions through SQL comments:

dbt\_project.yml

```yaml
models:
  PROJECT_NAME:
    +persist_docs:
      relation: true
      columns: true
```

## Python models

Python models run inside the dbt process and support the `table` materialization. `dbt.ref()` and `dbt.source()` return Arrow tables, loading each input into the process's memory. The `session` argument exposes the connection for SQL execution. Return an Arrow table or record batch reader, a pandas or polars DataFrame, or an object implementing the Arrow stream interface.

Incremental Python models are not supported.

## Relation names and access control

SQL relations render as `schema.table`, without the `redis` catalog. By default, the adapter doesn't quote schema or table names, and the driver folds unquoted identifiers to lowercase, as PostgreSQL does. Double-quoted names keep their exact case. To keep a name's case, enable dbt's `quoting` configuration for it.

Relations that driver releases earlier than 0.0.16 created with mixed-case names keep those names. Quote their exact names when you reference them, or rename them.

Redis controls access through ACL users, commands, and key patterns. The adapter skips SQL `grants` configurations with a warning. Configure permissions in Redis instead, following the [driver's ACL documentation](https://github.com/redis/redis-adbc-driver/blob/main/docs/driver-reference.md#acl-users).

## Transactions and interrupted runs

Every statement autocommits; the driver does not provide SQL transactions or rollback. A failed or cancelled materialization can leave partial writes and `__dbt_tmp` or `__dbt_backup` relations. dbt drops those relations at the start of the next run of that model. Account for this behavior when planning recovery from an interrupted run.

For package-specific macro overrides, refer to the adapter repository's [package compatibility documentation](https://github.com/redis/redis-dbt-adapter/blob/main/v1/README.md#package-compatibility).