Configure LLM providers

LLM (Large Language Model) providers power AI-driven features in DCT, including the AI Assistant and LLM-based data discovery. DCT supports a "Bring Your Own Provider" (BYOP) model, allowing you to connect your own LLM provider instead of using only Delphix-distributed models.

Supported provider types include: Embedded, Hosted Ollama, OpenAI, and Anthropic models. Each provider type has different configuration requirements.

Prerequisites

  • DCT Enterprise license tier

  • Administrator permissions

  • Access to LLM provider credentials or endpoints (varies by provider type)

Common configuration steps

All provider types follow these general steps:

  1. Navigate to Admin > AI Settings > Providers.

  2. Click Add Provider.

  3. Enter a Name for this provider.

  4. Select the Provider Type from the dropdown menu.

  5. Fill in the provider-specific configuration fields (see provider type sections below).

  6. Click Load Models (if applicable for your provider type).

  7. Select a Model from the available list.

  8. Optionally, select Set as default provider checkbox to use this provider for all LLM-powered features by default.

  9. Click Create.

Provider types

Anthropic

Anthropic provides a commercial LLM service accessible via API.

Configuration:

  • API Key (required) — Your Anthropic API key. Obtain this from your Anthropic account dashboard.

  • Custom Endpoint URL (optional) — Leave blank to use Anthropic's default public endpoint. No custom configuration is needed for standard Anthropic service.

Steps:

  1. Paste your API key in the API Key field.

  2. Click Load Models.

  3. Select a model from the list provided by Anthropic.

  4. Click Save.

OpenAI

OpenAI provides a commercial LLM service (e.g., GPT models) and can also be used with private or custom OpenAI-compatible instances.

Configuration:

  • API Key (required) — Your OpenAI API key. Obtain this from your OpenAI account.

  • Custom Endpoint URL (optional) — The OpenAI-compatible endpoint URL. Leave blank to use OpenAI's default public endpoint. Provide a custom endpoint if you are using a private OpenAI instance or an OpenAI-compatible service.

Steps:

  1. Paste your API key in the API Key field.

  2. If you are using a custom endpoint, enter the URL in the Custom Endpoint URL field.

  3. Click Load Models.

  4. Select a model from the list.

  5. Click Create.

Hosted Ollama

Hosted Ollama run LLM instances on your company's network and expose OpenAI-compatible endpoints. This provider type is also used for any other OpenAI-compatible LLM service.

Prerequisites:

  • An Ollama (or compatible) instance must be running on your company's network and configured to expose the OpenAI-compatible endpoint.

  • The endpoint must be accessible from the DCT instance.

Configuration:

  • Custom Endpoint URL (required) — The OpenAI-compatible endpoint URL, including the `/v1` path suffix. Example: http://host.docker.internal:11434/v1

Steps:

  1. Ensure your Ollama or compatible service is running and configured to expose the network endpoint.

  2. Obtain the OpenAI-compatible endpoint URL (must include `/v1` suffix).

  3. Enter the endpoint URL in the Custom Endpoint URL field.

  4. Click Load Models.

  5. Select a model from the list of available models on your Ollama instance.

  6. Click Create.

Embedded Model (Delphix-distributed)

Delphix provides pre-packaged, optimized language models that run locally within your DCT deployment. These models require no external API calls or network access to external services.

Currently, Embedded Models work only with the Secure Lookup feature in Continuous Compliance.

Prerequisites:

Configuration:

  • Model Upload (required) — Click to select and upload the model file from your local system.

Steps:

  1. Navigate to Delphix Download Center and download the file.

  2. In Admin > AI Settings > Providers, click Add Provider.

  3. Select Embedded as the Provider Type.

  4. Click the file picker to select your downloaded model file.

  5. Click Create. The upload process will begin.

  6. Any previously uploaded model will be replaced.

See Configuring AI services for more details on system requirements and deployment considerations.

Using a configured provider

Default provider

When you configure a provider and mark it as Set as default provider, it is automatically used for all LLM-powered features:

  • AI Assistant (when enabled)

  • LLM-based data discovery (when enabled)

Manual assignment

If you do not set a provider as default, you must manually assign it to specific LLM-powered features in Admin > AI Settings > LLM Features.

Activating LLM features

After you configure a provider:

  1. Navigate to Admin > AI Settings > LLM Features.

  2. Enable the desired LLM-powered feature (e.g., AI Assistant, Data Classification).

  3. If you set a default provider, it will be automatically selected.

The LLM-powered feature is now active and available in Synthetic Data.