AI Provider Configuration Reference
This topic lists every field in the Provider configuration section of the Create AI provider and Edit AI provider forms, grouped by provider type.
For the steps to add a provider, see Manage AI Providers.
Each form shows the essential fields first and keeps the rest under Additional Configuration Options, so a typical connection needs only the required fields. Fields that hold a secret, such as an API key or an AWS key, appear as masked inputs.
OpenAI
| Field | Required | Description |
|---|---|---|
api_key | Yes | OpenAI API key for authentication. |
base_url | No | Custom base URL for the OpenAI API endpoint. |
organization | No | OpenAI organization ID to use for requests. |
project | No | OpenAI project ID, to scope requests to a specific project. |
webhook_secret | No | Secret for validating OpenAI webhooks payloads. |
websocket_base_url | No | Custom base URL for OpenAI WebSocket connections. Defaults to the standard OpenAI endpoint. |
Anthropic
| Field | Required | Description |
|---|---|---|
api_key | Yes | Anthropic API key for authentication. |
base_url | No | Custom base URL for the Anthropic API endpoint. |
timeout | No | Timeout in seconds for API requests. Defaults to 60. |
max_retries | No | Maximum number of retries for failed API requests. Defaults to 2. |
default_headers | No | Additional HTTP headers to include with every API request. |
anthropic_proxy | No | Proxy URL for Anthropic API HTTP connections. |
AWS Bedrock
| Field | Required | Description |
|---|---|---|
region_name | Yes | AWS region to connect to, such as us-west-2. A provider connects to a single region. |
aws_access_key_id | Yes | AWS access key ID. Provide it together with the secret access key. |
aws_secret_access_key | Yes | AWS secret access key. Provide it together with the access key ID. |
aws_session_token | No | AWS session token, for temporary credentials. Requires both the access key ID and the secret access key. |
aws_account_id | No | AWS account ID to use when creating the client. |
endpoint_url | No | A complete URL, including the scheme, that overrides the automatically constructed endpoint. When set, use_ssl is ignored. |
api_version | No | Boto3 API version to use. Leave it unset to use the latest. |
batch_s3_bucket | No | S3 bucket name for batch inference input and output. Required for batch operations. |
batch_role_arn | No | Amazon Resource Name (ARN) of the IAM role that Bedrock assumes to reach S3 for batch inference. |
config | No | Advanced client options. Only connect_timeout and read_timeout, both in seconds, are supported. |
use_ssl | No | Whether to use SSL. On by default. |
verify | No | SSL certificate verification. Set it to false to skip verification, or provide a path to a certificate authority (CA) bundle. |
Although the AWS credential fields are individually optional, they're validated as a set:
- The access key ID and the secret access key must be provided together.
- A session token requires both the access key ID and the secret access key.
In the form, the region and the AWS keys appear as the primary fields, and everything else sits under Additional Configuration Options. The batch fields matter only if you run batch inference.