Skip to main content

​
Bring Your Own Cloud

This guide provides step-by-step instructions for deploying the Collate AI Proxy (CAIP) with Helm alongside your existing Collate installation. The chart packages the Collate AI Proxy as a standalone Kubernetes service that integrates with Collate.
Collate AI requires the underlying Collate deployment to use OpenSearch as its search engine. ElasticSearch is not supported because Semantic and Hybrid Search rely on OpenSearch’s vector capabilities.

​
Architecture

The following diagram shows the deployment architecture for the Collate AI Proxy when it is installed with Helm alongside an existing Collate deployment. In AWS, we rely on AWS Bedrock using its regional endpoints or inference endpoints to interact with large language models and power our AI agents. For other providers, such as Azure OpenAI, we query the provider APIs directly.

​
Guide sections

Prerequisites

Review the software, access, hardware, and Azure OpenAI setup requirements.

Deployment

Configure Helm values and install the Collate AI Proxy.

Configure Collate

Connect your Collate deployment to the Collate AI Proxy.

Validation

Verify that the deployment is healthy from the Collate UI.

Configuration reference

Configure bot authentication, model tiers, context retrieval, logging, and error reporting.

Additional settings

Override image settings and host your own Docker images if needed.