Split Agentic chat to its own feature setting
What does this MR do and why?
This MR splits "Agentic chat" to its own feature setting.
| Before | After |
|---|---|
|
|
This change introduces a new AI feature setting called "Agentic Chat" that's separate from the existing "Duo Agent Platform" feature. Previously, agentic chat functionality was bundled together with other agent platform features under a single setting (feature ID 16). Now it gets its own dedicated setting (feature ID 17).
The update includes three database migration scripts that copy existing agent platform configurations to the new agentic chat feature for different deployment types (self-hosted, self-managed, and GitLab.com). This ensures existing users don't lose their settings when the feature is split.
A feature flag called ai_agentic_chat_feature_setting_split controls this change - when enabled, the system uses the new separate agentic chat setting; when disabled, it falls back to the old combined approach. When disabled, this also ensures that the setting for "Agentic chat" configuration does not appear in the model selection UI. This allows for a gradual rollout and easy rollback if needed.
The change also updates documentation, test files, and configuration to recognize the new feature setting. The main benefit is that administrators can now configure agentic chat independently from other agent platform features, providing more granular control over AI functionality.
In the AI Gateway, models for duo_agent_platform_agentic_chat has already been added: gitlab-org/modelops/applied-ml/code-suggestions/ai-assist@c1619bc8
Migration log
https://gitlab.com/gitlab-org/gitlab/-/jobs/12319576491
References
Screenshots or screen recordings
| Before | After |
|---|---|
How to set up and validate locally
MR acceptance checklist
Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.

