Database reference

Each workspace has an ingestion PostgreSQL database. The public query API uses your workspace API key. Direct PostgreSQL access uses a separate database role. For connection and query procedures, see Query your Database.

Schemas

Schema	Contract
`enrichments`	Deepline manages enrichment data, including `enrichments.enrichments` and `enrichments.companies`. Customers can read this data but cannot change it.
`crm`	Deepline manages normalized CRM and sequencing data. Customers can read this data but cannot change it.
Provider schemas	Managed data for `hubspot`, `salesforce`, `attio`, `instantly`, `lemlist`, `heyreach`, `smartlead`, and `theirstack`. Available rows depend on configured ingestion.
Monitor schemas	Managed data for `deepline_native`, `parallel`, `firecrawl`, `findymail`, and `predictleads`. Available rows depend on configured Monitors.
`dl_catalog`	Managed object-materialization metadata.
`dl_ingest`, `dl_meta`	Deepline manages ingestion and metadata objects. The database prevents customer writes to these objects.
`storage`	Customer storage contains files and customer-created tables. Separate permissions protect internal runtime receipt objects.
Customer-managed schemas	Customers can create schemas with names outside the reserved set. Schema-qualified SQL can create and change objects in these schemas.

The bootstrap does not provide dl_resolved.people, dl_graph, dl_cache, or dl_override as a managed identity-graph contract.

SQL API

POST /api/v2/db/query accepts one SQL statement and max_rows (up to 1,000). The response contains rows, columns, row_count_returned, and truncated. The SDK equivalent is client.db.query({ sql, maxRows }). Reads include SELECT, read-only WITH, and EXPLAIN subject to the query policy. Writes must identify a schema-qualified target in storage or an allowed customer-managed schema. Managed data, internal runtime objects, public, information_schema, and pg_* schemas are not customer write targets. Retired internal schema names are also reserved. Choose a new schema name when the query API reports a reserved target. Example against a provisioned company table:

SELECT domain, company_name, industry, employee_count
FROM enrichments.companies
ORDER BY updated_at DESC
LIMIT 20;

Provisioning creates this table. The table can contain no rows.

Roles and direct connections

dl_ingest_customer_db: This customer connection role reads managed tables. Ownership grants let this role own permitted customer storage objects.
dl_ingest_runtime: This Deepline runtime role can write managed ingestion data and customer storage.
dl_ingest_customer_owner: ownership role for customer-managed objects.
dl_ingest_owner, dl_ingest_storage_owner, and dl_ingest_protected_owner: platform ownership roles, not customer credentials.

URI reveal, provisioning, and password rotation require the dashboard/admin session flow. A workspace API key for SQL queries is not a URI-reveal credential.

Persistence and caching

Saved database rows and reusable tool-call results serve different purposes. Persistence alone does not make future provider requests free. The SDK freshness options control durable call reuse:

`staleAfterSeconds`	Behavior
`null`	Allows indefinite reuse of a completed result.
`0`	Starts fresh logical work; a completed operation receipt can recover the same operation after resume.
Positive integer	Sets the maximum age of a reusable result, in seconds.

Edition 11 requires an explicit choice. Earlier admitted editions retain their omitted-value behavior. The admitted authoring contract determines the applicable rules. Reuse also depends on the operation, inputs, authentication identity, and version policy. Certain live-state tools are always fresh. Plays execute at least once. Cache reuse does not guarantee that a provider effect occurs exactly once.

​Schemas

​SQL API

​Roles and direct connections

​Persistence and caching

Schemas

SQL API

Roles and direct connections

Persistence and caching