By default, users can access the AI assistant by using the floating Ask AI button in the bottom-right corner or through the Search modal in the top navigation bar.
Use the aiAssistant configuration to:
- hide the AI assistant
- add suggested pages to the search modal
- set the built-in prompt text
- add an Ask AI button
The aiAssistant option also supports page-level configuration using front matter.
The AI assistant indexes and searches across all content in your project:
- API descriptions - OpenAPI, GraphQL, AsyncAPI, and SOAP description files
- Documentation pages - Markdown files
- API versions - All versions of APIs are indexed and filtered based on the active version
When you have multiple versions of an API, the AI assistant shows results from the active version by default. Results also include default versions from other APIs and non-versioned content.
| Option | Type | Description |
|---|---|---|
| hide | boolean | Hides the AI assistant when set to true. Default: false. |
| prompt | string | Built-in instructions for the AI assistant. Applied to all AI assistant answers in the project and not visible to users. Use to set greeting, tone, or other answer conditions. |
| suggestions | [string] | List of suggestions displayed in the AI assistant interface. |
| trigger | Trigger button | Configures the Ask AI floating button. |
| Option | Type | Description |
|---|---|---|
| hide | boolean | Hides the trigger button when set to true. Default: false. |
| inputType | string | Type of UI component used to render trigger button. Possible values: button, icon. Default: button. |
| inputIcon | string | Icon of the trigger button component. Possible values: redocly, sparkles, chat. Default: sparkles. |
The AI assistant is available on Enterprise and Enterprise Plus plans with a monthly limit of 3500 questions per organization. The limit resets at the beginning of each month. Above the limit, users receive regular search results instead of AI answers. These limits can change.
Data usage and privacy: Curious how the AI assistant uses your data? The Redocly AI assistant runs in inference-only mode and does not train or fine-tune AI models on your content. For details, see the AI governance and security FAQ.
AI assistant and Typesense search indexes are only built on the production branch. Changes to search configuration or content exclusions, like the excludeFromSearch front matter option, may not immediately appear in search results until the next production build.
Display the AI assistant button with a custom prompt:
aiAssistant:
hide: false
prompt: Speak only in rhymesSet AI assistant suggestions:
aiAssistant:
hide: false
suggestions:
- How to create a new API?
- What is Redocly?
- How to manage an organization?
aiAssistant:
trigger:
hide: false
inputType: button
inputIcon: redocly- AI governance and security FAQ - How the AI assistant works, what data it uses, and which controls protect it
search- Learn how to customize search in your project, curate search results, and exclude pages from search- Configuration options - Explore other project configuration options for comprehensive documentation and platform customization