Documentation Index
Fetch the complete documentation index at: https://docs.superblocks.com/llms.txt
Use this file to discover all available pages before exploring further.
Overview
By default, Clark AI runs inference through models hosted in the Superblocks Cloud. With Bring Your Own Inference, administrators can redirect Clark to run against their organization’s own AI infrastructure instead.Bring Your Own Inference is available on Cloud-Prem deployments only.
Supported providers
| Provider | Description |
|---|---|
| Anthropic API | Direct API access to Anthropic models |
| AWS Bedrock | Managed inference through Amazon Bedrock |
| Custom AI Gateway | Route through your own AI gateway for centralized policy, logging, and model management |
| Databricks AI Gateway | Inference routed through a Databricks-managed AI Gateway |
| Google Vertex AI | Managed inference through Google Cloud Vertex AI |
| Snowflake Cortex | Inference through Snowflake’s built-in Cortex AI service |
How it works
1. Admin configures the inference provider
In Organization Settings, an administrator selects their inference provider and supplies the necessary credentials — for example, an API endpoint and authentication token, or a personal access token for AWS Bedrock.2. All builders inherit the configuration
Once saved, every builder in the organization automatically uses the configured provider for all Clark interactions. There is no per-user or per-app setup required.3. Inference burns down your commitment and reduces Clark credit usage
Every prompt a builder sends — whether exploring a database schema, generating a plan, or writing code — runs inference on your provider. Token consumption counts toward your existing cloud commitment and reduces Clark AI credit usage.What this means for your organization
- Burn down your existing commitment: Inference runs on your provider, contributing to your cloud commitment and reducing Clark credit consumption
- Inference runs in your network: For providers like Bedrock and Vertex, all model inference executes within your cloud account
- No builder friction: Builders use Clark identically regardless of the underlying provider