canonical: https://jentic.com/apis/googleapis.com/datastream

# Google Datastream API

Google Cloud Datastream is a serverless change-data-capture (CDC) and replication service that streams changes from operational databases (Oracle, MySQL, PostgreSQL, SQL Server) into BigQuery, Cloud Storage, and other destinations. The API exposes 21 endpoints covering connection profiles for source and destination, streams, private connectivity for VPC peering, and routes. It supports continuous replication with backfill, schema discovery, and per-stream pause/resume.

## For AI agents

Set up and operate change-data-capture streams from Oracle, MySQL, PostgreSQL, and SQL Server into BigQuery and Cloud Storage. Lets agents drive CDC pipelines without standing up custom replication.

## Scope

Does not run database queries, transform data in flight beyond per-row mapping, or replicate to non-Google destinations - use for serverless CDC from operational databases into BigQuery and Cloud Storage only.

## Capabilities

- Create source connection profiles for Oracle, MySQL, PostgreSQL, and SQL Server
- Create destination connection profiles for BigQuery and Cloud Storage
- Run CDC streams with optional historical backfill on tables
- Pause and resume streams to control downstream load
- Discover source schema and tables before starting a stream
- Configure private connectivity and routes for VPC-peered sources

## Use cases

### Operational-Database to BigQuery CDC

Stream changes from Oracle, MySQL, PostgreSQL, or SQL Server into BigQuery with minutes of latency, plus an initial backfill of historical rows. The API creates source and destination connection profiles, then a stream that reads source binlogs/redo logs and lands change events in BigQuery tables. Setup is typically hours rather than the weeks needed for a custom CDC pipeline.

Example prompt: Create source profile 'mysql-prod', destination profile 'bq-warehouse', and stream 'orders-cdc' with backfillAll on; start the stream and report rowsBackfilled per object

### Schema Discovery Before Stream Setup

Before launching a stream, call discoverConnectionProfile on the source profile to enumerate available schemas, tables, and columns. The response surfaces primary keys and unsupported types, so the operator can decide which tables to include and whether to add type-mapping overrides. This avoids stream creation failures from invalid include lists.

Example prompt: Call discoverConnectionProfile on 'mysql-prod' with hierarchyDepth 2, then return the full list of schemas and tables to the operator for selection

### Backfill and Pause-Resume Operations

Trigger a backfill on individual stream objects after a schema change, and pause and resume streams to coordinate with downstream maintenance. POST /v1/{+object}:startBackfillJob and :stopBackfillJob give per-table control without restarting the whole stream. Pause/resume on the parent stream is exposed via the patch endpoint with the state field.

Example prompt: Call startBackfillJob on stream object 'orders-cdc/orders' to replay history after adding a column, then resume CDC on the parent stream

### AI Agent CDC Operator

An AI agent can stand up and operate Datastream pipelines through Jentic without operator-written replication code. Jentic search returns the matching connection-profile, stream, or backfill operation, the agent loads the schema, and Jentic executes against datastream.googleapis.com using vault-stored credentials. This compresses the multi-day setup of OAuth and Datastream resources into one agent run.

Example prompt: Use Jentic to search 'create a datastream stream', load the createStream schema, and execute it with the configured source and destination profiles plus an includeObjects table list

## Key endpoints

| Method | Path | Description |
| --- | --- | --- |
| POST | /v1/{+parent}/connectionProfiles | Create a source or destination connection profile |
| POST | /v1/{+parent}/streams | Create a Datastream stream |
| POST | /v1/{+name}:run | Start a stream |
| POST | /v1/{+object}:startBackfillJob | Start backfill on a stream object |
| POST | /v1/{+object}:stopBackfillJob | Stop backfill on a stream object |
| GET | /v1/{+name}:fetchStaticIps | Fetch Datastream egress IPs for source allowlisting |

## Key resources

- **ConnectionProfiles** — Source and destination database/storage connection settings
- **Streams** — Continuous CDC streams with optional backfill
- **PrivateConnections** — VPC-peered private connectivity from Google to source networks
- **Routes** — Routes that direct traffic over private connections
- **StreamObjects** — Per-table configuration and backfill control inside a stream

## Why Jentic

- **Setup:** Wiring the Datastream API by hand means setting up Google OAuth, refreshing short-lived scoped tokens instead of holding client secrets in code, and addressing connection profile and stream resources on datastream.googleapis.com. Through Jentic you install once, import the Datastream API from the API Directory, store the Google credential once, and your agent calls it.
- **Permission scoping:** The API carries the parent and resource name in the URL path (/v1/{parent}/streams and /v1/{name}:run), so a rule can pin your agent to one stream or connection profile. You choose the operations it may call, so it can create and run a stream and start a backfill while stopping a backfill is not included unless you add it.
- **Credential handling:** Your Google OAuth credential is stored once, encrypted, by your own Jentic One instance and injected at execution time. It never enters the agent's prompt, logs, or context.
- **Discovery method:** Agents search Jentic by intent such as 'create a Datastream stream', and Jentic returns the matching Datastream operation with its input schema so the agent calls the right endpoint without browsing the reference docs.

## Related APIs

- **Database Migration API** — Database Migration is for one-time database moves with cutover; Datastream is for ongoing CDC.
- **BigQuery API** — BigQuery is Datastream's primary destination; the BigQuery API queries the captured tables.
- **Cloud Pub/Sub API** — Pub/Sub provides general event streaming; Datastream is purpose-built for database CDC.

## FAQ

### What authentication does the Datastream API use?

The Datastream API uses OAuth 2.0 with the cloud-platform scope. Through Jentic the OAuth client and refresh tokens are stored in your Jentic One instance and the agent receives short-lived scoped access tokens, so raw Google credentials never enter the agent context.

### Can I stream Oracle changes to BigQuery with Datastream?

Yes. Create a source connection profile of type ORACLE, a destination connection profile of type BIGQUERY, then a stream that references both. Datastream uses Oracle LogMiner to capture changes and writes per-table change events to BigQuery with seconds-to-minutes lag.

### What are the rate limits for the Datastream API?

Google enforces standard Cloud quotas on datastream.googleapis.com: per-project rate limits on read/write calls and per-region quotas on concurrent streams and connection profiles. Quotas are visible in the Cloud Console under IAM and admin, quotas, filtered to datastream.googleapis.com.

### How do I allowlist Datastream's source IPs?

Call GET /v1/{+name}:fetchStaticIps on the location resource. The response returns the egress IP ranges Datastream uses for IP-allowlist source connectivity in that region; add those to the source database's network allowlist before creating the stream.

### Is the Datastream API free?

API calls are free. Datastream is billed per GiB of changed data captured plus a small per-stream operating fee. Underlying BigQuery storage and Cloud Storage destinations are billed at standard rates.

### How do I backfill a single table after a schema change?

Call POST /v1/{+object}:startBackfillJob on the stream object that represents the table. Datastream replays historical rows for that object only, leaving CDC running on the rest of the stream. Stop a runaway backfill with :stopBackfillJob.

### Can I limit what my agent is allowed to do with the Datastream API?

Yes. Because you run Jentic One yourself, your own rules decide which Datastream operations and which Google credential the agent may use. Since the parent and resource name travel in the URL path, such as /v1/{parent}/streams and /v1/{name}:run, you can pin the agent to a single stream or connection profile. You might allow it to create and run a stream and start a backfill while withholding stopBackfillJob unless you explicitly add it.