# Use SQL views as input tables

Use a SQL view when you need to reshape data *before* Syntho reads it.

This is most common for:

* **AI synthesize** workflows that work best on a single entity table.
* **Data sharing** where recipients prefer one clean table.
* **Analytics sandboxes** where BI dashboards query a flattened model.

### Why this matters

A well-designed input table makes configuration simpler and safer:

* Fewer joins means fewer linkage points.
* You can enforce “one row = one entity” rules.
* You can precompute derived fields once, then keep Syntho configs minimal.

### Practical pattern: create a single entity view

1. Pick the entity you care about (customer, patient, account).
2. Join only what you need for the use case.
3. Ensure every row has a stable, unique key.

{% hint style="info" %}
Syntho can read from a **table or view** in the source. See [Table view](https://docs.syntho.ai/configure-a-data-generation-job/configure-table-settings).
{% endhint %}

#### Example (PostgreSQL): flatten customers + last order

```sql
CREATE OR REPLACE VIEW syntho_customer_entity_v AS
SELECT
  c.customer_id,
  c.country_code,
  c.segment,
  c.signup_date,
  o.order_date  AS last_order_date,
  o.order_total AS last_order_total
FROM customers c
LEFT JOIN LATERAL (
  SELECT order_date, order_total
  FROM orders o
  WHERE o.customer_id = c.customer_id
  ORDER BY order_date DESC
  LIMIT 1
) o ON true;
```

### Destination considerations

Syntho writes to the **destination**. The destination object must usually be a **table**.

If your source is a view:

* Create a destination table with the same columns.
* Or write to a filesystem destination (e.g., Parquet), if that fits your workflow.

### Edge cases and gotchas

* **Missing primary keys**: add a stable surrogate key in the view, or model a different entity.
* **Exploding joins** (1-to-many): aggregate first (e.g., counts, sums, “last event”), then join.
* **Materialized views**: use them when the view becomes slow to query during training.
* **Security**: grant **read-only** access to the view, not the full underlying schema.

### Where to use this in Syntho

* Use the view as the source table when configuring [AI synthesize](https://docs.syntho.ai/configure-a-data-generation-job/configure-column-settings/ai-powered-generation).
* Keep the generator config small. Prefer calculated columns for only the rules you must enforce.