Skip to main content
Only Volcengine is currently activated in the codebase. The environment schema preserves fields for other providers (OpenAI, Anthropic, Azure OpenAI, Vercel, Moonshot), but none of those are wired up in apps/server/src/config/ai-providers.ts. Set only VOLCENGINE_API_KEY and VOLCENGINE_ENDPOINT for AI reviews to function.

Overview

AI Review uses a configurable AI provider system. Each provider exposes one or more models. Models can be defined at the global level or overridden per project. The active provider list is controlled in apps/server/src/config/ai-providers.ts:

Environment variables

Set these two variables before starting the server:

Preserved but inactive variables

The following variables are parsed by the server’s environment schema but have no active provider implementation: Setting these variables has no effect on review execution in the current release.

Configure AI in the dashboard

1

Open AI Configuration

Log in to the dashboard and navigate to Settings → AI Configuration.
2

Add an AI configuration

Click Add Configuration and fill in the fields:
3

Set as default (optional)

Mark the configuration as the global default so all projects without a per-project override use it.
4

Verify

Use the test button to confirm the provider responds before saving.

Per-project AI configuration

Each project can override the global default AI configuration:
  1. Open the project from the Projects page.
  2. Go to the project’s Settings tab.
  3. Under AI Configuration, select a specific AI configuration or leave it set to Use global default.
When a review is triggered for a project, the server resolves configuration in this order:
  1. Per-project AI configuration (if set)
  2. Global default AI configuration

Managing AI models

AI models are managed separately from AI configurations and represent the specific model identifiers available for selection. To add or manage models:
  1. Navigate to Settings → AI Models in the dashboard.
  2. Click Add Model to register a new model ID.
  3. Assign the model to an AI configuration when creating or editing that configuration.
Model IDs must match the identifiers expected by the Volcengine endpoint. Check your Volcengine console for the exact model endpoint names.