> For the complete documentation index, see [llms.txt](https://docs.algenta.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.algenta.ai/connectors/gcs.md).

# Google Cloud Storage

By the end of this page you will have a saved `gcs` connector that reaches a bucket, a confirmed `live` access check, a browse listing of the objects it can see, one of those objects onboarded as a queryable **dataset**, and a first [governed query](/guides/governed-query.md) returning rows from it. The engine reads the object in place, parses it (CSV, TSV, JSON, Parquet, or Excel), and registers the rows as [governed data](/concepts/governed-data.md) under a typed, validated contract.

Prerequisites: an API key and a base URL (see [Authentication](/getting-started/authentication.md)), and access to a GCS bucket — either a service account or a public bucket. Use `https://api.algenta.ai` for Algenta cloud, or your own origin such as `http://localhost:8000` for a self-hosted deployment.

{% hint style="info" %}
A `gcs` connector reads these `config` fields: `bucket` and `blob` (the object path) identify the object and are both required to test the connection; `file_type` overrides the format otherwise inferred from the blob's extension; `json_path` is a dot-path to a records array inside a JSON file; `project_id` sets the billing project. For credentials, supply `credentials` (a path to a service account JSON file on the engine host, or an inline JSON string), or `service_account_json` (an inline JSON string), or set `anonymous: true` to read a public bucket without credentials.
{% endhint %}

{% stepper %}
{% step %}

### Set your credentials

Export your API key and base URL. If you authenticate with a service account JSON, keep it in its own variable.

```bash
export ALGENTA_API_KEY="$ALGENTA_API_KEY"
export ALGENTA_BASE_URL="https://api.algenta.ai"
export GCP_SERVICE_ACCOUNT_JSON="$GCP_SERVICE_ACCOUNT_JSON"
```

Verify the variables are set:

```bash
echo "${ALGENTA_API_KEY:?set ALGENTA_API_KEY}" >/dev/null && echo "key present"
echo "$ALGENTA_BASE_URL"
```

**Expected result:** `key present` and your base URL print without error. If either errors, export the missing variable before continuing.
{% endstep %}

{% step %}

### Confirm the engine accepts `gcs`

Ask the engine which connector types it supports and confirm `gcs` is in the list.

```bash
curl -s "$ALGENTA_BASE_URL/v1/connectors/types" \
  -H "Authorization: Bearer $ALGENTA_API_KEY"
```

**Expected result:** a JSON object with a `types` array that includes `"gcs"`, for example `{"types": ["azure", "gcs", "s3", ...]}`.
{% endstep %}

{% step %}

### Create the GCS connector

`POST /v1/connectors` stores the connection name, type, and credentials, and returns the connector with `status: "untested"`. Set `bucket`, `blob`, and your credentials. This example reads a public bucket with `anonymous`:

```bash
curl -s -X POST "$ALGENTA_BASE_URL/v1/connectors" \
  -H "Authorization: Bearer $ALGENTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"name\": \"Public GCS Data\",
    \"connector_type\": \"gcs\",
    \"config\": {
      \"bucket\": \"open-datasets\",
      \"blob\": \"census/2020.parquet\",
      \"anonymous\": true
    }
  }"
```

For a private bucket on a self-hosted engine, point `credentials` at a service account JSON file the engine can read, and set `project_id`:

```bash
curl -s -X POST "$ALGENTA_BASE_URL/v1/connectors" \
  -H "Authorization: Bearer $ALGENTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"name\": \"Analytics GCS\",
    \"connector_type\": \"gcs\",
    \"config\": {
      \"bucket\": \"acme-analytics\",
      \"blob\": \"exports/orders.csv\",
      \"project_id\": \"acme-prod\",
      \"credentials\": \"/secrets/gcp-sa.json\"
    }
  }"
```

Capture the returned `id` for the steps that follow:

```bash
export CONNECTOR_ID="$CONNECTOR_ID"
```

**Expected result:** a `201` response with the connector `id`, `"connector_type": "gcs"`, and `"status": "untested"`. Credentials are encrypted at rest.

{% hint style="warning" %}
The connection test reads both `bucket` and `blob`; if `blob` is missing the test fails. Point `blob` at a real object (for example the one you plan to onboard). On Algenta cloud, prefer `service_account_json` (an inline JSON string) or `anonymous`, since a file path in `credentials` must exist on the engine host.
{% endhint %}
{% endstep %}

{% step %}

### Test object access

`POST /v1/connectors/{id}/test` makes a real request that confirms `gs://bucket/blob` is readable, then flips the connector to `live` on success or `error` on failure.

```bash
curl -s -X POST "$ALGENTA_BASE_URL/v1/connectors/$CONNECTOR_ID/test" \
  -H "Authorization: Bearer $ALGENTA_API_KEY"
```

**Expected result:** `{"success": true, "status": "live", "latency_ms": 240, "message": "gs://acme-analytics/exports/orders.csv accessible."}`, and the connector's stored `status` is now `live`. On failure, `error_type` (for example `not_found`, `auth`, or `permission`) plus `recoverable` guide your fix.
{% endstep %}

{% step %}

### Browse the bucket

`GET /v1/connectors/{id}/browse` lists up to 100 objects in the bucket, so you can pick what to onboard. This returns `409 not_connected` if the connector is not `live` — run the test step first, and make sure `bucket` is set.

```bash
curl -s "$ALGENTA_BASE_URL/v1/connectors/$CONNECTOR_ID/browse" \
  -H "Authorization: Bearer $ALGENTA_API_KEY"
```

**Expected result:** a response with `"connector_type": "gcs"`, an `items` array (each entry is an object, with its `selection.object_path`), a `total`, and a `message` such as `"Found 19 objects in gs://acme-analytics"`. Note the `object_path` you want to onboard — for example `exports/orders.csv`.
{% endstep %}

{% step %}

### Onboard one object as a dataset

`POST /v1/data/connect` ties the saved connector to a selection — an object path — and registers the parsed rows as a queryable dataset. Reference the connector by `connection_id` and name the object in `selection.object_path`.

```bash
curl -s -X POST "$ALGENTA_BASE_URL/v1/data/connect" \
  -H "Authorization: Bearer $ALGENTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"dataset_name\": \"sales_orders\",
    \"connection_id\": \"$CONNECTOR_ID\",
    \"selection\": {\"object_path\": \"exports/orders.csv\"},
    \"visibility\": \"private\"
  }"
```

The format is inferred from the object's extension; override it with `file_type` in the connector `config` (one of `csv`, `tsv`, `json`, `parquet`, `excel`), and use `json_path` to point at a records array inside a JSON file.

**Expected result:** `"status": "ready"` with a `dataset_id`, a `schema_summary` (row and column counts), and the `connection_id`. If more than one object matches and no selection is given, you get `"status": "needs_selection"` with a `choices` array — re-call with one `object_path` from `choices`. Save the `dataset_id`:

```bash
export DATASET_ID="$DATASET_ID"
```

{% endstep %}

{% step %}

### Run a governed query

The parsed object is now governed data. `POST /v1/query` runs a deterministic, typed query against the dataset and returns rows under the engine's validated contract.

```bash
curl -s -X POST "$ALGENTA_BASE_URL/v1/query" \
  -H "Authorization: Bearer $ALGENTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"dataset_id\": \"$DATASET_ID\",
    \"select\": [\"region\", \"amount\"],
    \"limit\": 10
  }"
```

**Expected result:** a `200` response with the matching rows and the resolved query plan. Re-running the same request against the same dataset returns the same rows in the same order — governed queries are deterministic by design.
{% endstep %}
{% endstepper %}

## Expected result

You have a saved `gcs` connector with `status: "live"`, a browse listing of a bucket's objects, a dataset onboarded from one of them that appears in `GET /v1/data` with a `dataset_id` and a profiled schema, and a first query returning rows. The same intent against the same dataset returns the same deterministic answer, and every access is auditable.

## Other ways to connect

The same endpoints back the `de` CLI and the Python SDK — all three hit the identical API.

```bash
de connectors create connector.json        # POST /v1/connectors from a JSON body
de connectors test $CONNECTOR_ID           # real object-access check
de connectors browse $CONNECTOR_ID         # browse the bucket's objects
de data connect --request connect.json     # POST /v1/data/connect from a JSON body
de data list                               # list registered datasets
```

```python
import os
from decision_engine import AlgentaClient

client = AlgentaClient(
    api_key=os.environ["ALGENTA_API_KEY"],
    base_url="https://api.algenta.ai",
)

connector = client.create_connector(
    name="Analytics GCS",
    connector_type="gcs",
    config={
        "bucket": "acme-analytics",
        "blob": "exports/orders.csv",
        "project_id": "acme-prod",
        "service_account_json": os.environ["GCP_SERVICE_ACCOUNT_JSON"],
    },
)

test = client.test_connector(connector.id)
assert test.success, test.message

result = client.connect_data(
    connection_type="object_storage",
    provider="gcs",
    dataset_name="sales_orders",
    connection_id=connector.id,
    selection={"object_path": "exports/orders.csv"},
    visibility="private",
)
print(result.status, result.dataset_id)
```

## Troubleshooting

<details>

<summary>Test fails because `blob` is not set</summary>

The GCS test reads both `bucket` and `blob`. Set `blob` to a real object path (typically the object you plan to onboard) and re-test.

</details>

<details>

<summary>Test fails with `error_type: "auth"` or `"not_found"`</summary>

For `auth`, the service account is missing or lacks access — check `credentials` / `service_account_json` and that the account can read the bucket. For `not_found`, confirm `bucket` and `blob` name a real object. A private bucket read anonymously will also fail.

</details>

<details>

<summary>Test fails with a privacy-policy or permission message</summary>

`error_type` is `permission`. Google Cloud Storage is outside the egress allowlist for your deployment. Confirm the destination is permitted by your deployment's egress policy. See [Troubleshooting](/help/troubleshooting.md).

</details>

<details>

<summary>`409 not_connected` when browsing</summary>

The connector is not `live`. Run `POST /v1/connectors/{id}/test` first; only a connector that passes its test can be browsed. Browsing also requires `bucket` to be set. Editing a connector's `config` resets it to `untested`, so re-test after any change.

</details>

## Next

{% content-ref url="/pages/1hWq4uaSOQ9pNlD9Ox0f" %}
[Query governed data](/guides/governed-query.md)
{% endcontent-ref %}

{% content-ref url="/pages/Pv08Fr0IvnPwOTKivI9D" %}
[Amazon S3](/connectors/s3.md)
{% endcontent-ref %}

{% content-ref url="/pages/mTceQEYGKRUXY86O0KHF" %}
[Azure Blob Storage](/connectors/azure-blob.md)
{% endcontent-ref %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.algenta.ai/connectors/gcs.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.