diff --git a/fern/products/dashboard/pages/overview.mdx b/fern/products/dashboard/pages/overview.mdx index 2e4a2baf3..e5d221d0e 100644 --- a/fern/products/dashboard/pages/overview.mdx +++ b/fern/products/dashboard/pages/overview.mdx @@ -191,6 +191,39 @@ Manage your documentation site and team collaboration: Temporarily take a site offline without deleting it. + +## AI features + +The dashboard uses AI in these places. Each runs only when you invoke it, and the page content, URL, or diff you provide is sent to a model along with your organization and site context. Fern doesn't use this data to train AI models. + + + + Classifies the pages of an existing documentation site into a proposed Fern navigation structure during onboarding. + + + Reads a public URL and proposes colors, typography, and logos for your docs site. + + + Generates a title and description for pull requests opened from Fern Editor. + + + Answers questions about your site, configuration, and analytics, and makes edits as a Fern Editor draft or pull request. + + @@ -228,5 +261,14 @@ Manage your documentation site and team collaboration: - [Password protection](/learn/dashboard/configuration/password-protection): Restrict access to your docs site with a shared password. - [Query log retention](/learn/docs/ai-features/ask-fern/overview#features): Purge Ask Fern end-user query logs on a schedule instead of keeping them indefinitely. - [Unpublish a site](/learn/docs/preview-publish/publishing-your-docs#unpublishing-your-docs): Temporarily take a site offline without deleting it. + +## AI features + +The dashboard uses AI in these places. Each runs only when you invoke it, and the page content, URL, or diff you provide is sent to a model along with your organization and site context. Fern doesn't use this data to train AI models. + +- **Site import**: Classifies the pages of an existing documentation site into a proposed Fern navigation structure during onboarding. +- **Brand extraction**: Reads a public URL and proposes colors, typography, and logos for your docs site. +- [Pull request summaries](/learn/docs/writing-content/fern-editor): Generates a title and description for pull requests opened from Fern Editor. +- [Fern Agent](/learn/docs/ai-features/fern-agent): Answers questions about your site, configuration, and analytics, and makes edits as a Fern Editor draft or pull request. diff --git a/fern/products/docs/pages/ai/ai-examples.mdx b/fern/products/docs/pages/ai/ai-examples.mdx index 2170e836b..344262c34 100644 --- a/fern/products/docs/pages/ai/ai-examples.mdx +++ b/fern/products/docs/pages/ai/ai-examples.mdx @@ -9,6 +9,8 @@ Fern uses AI to automatically generate realistic request and response examples f AI examples work alongside manual examples defined in [`x-fern-examples`](/learn/api-definitions/openapi/extensions/request-response-examples), with manual examples taking priority. +Generation sends endpoint metadata, the relevant part of your API definition, and existing examples to the model. Fern doesn't use your API definition to train AI models. + diff --git a/fern/products/docs/pages/ai/overview.mdx b/fern/products/docs/pages/ai/overview.mdx index 3dad9c562..192278b45 100644 --- a/fern/products/docs/pages/ai/overview.mdx +++ b/fern/products/docs/pages/ai/overview.mdx @@ -43,7 +43,9 @@ You can hand your docs to an agent. [Fern Agent](/learn/docs/ai-features/fern-ag ## Optimize for AI -Your site is automatically optimized for AI tools and search engines. Fern hosts an `llms.txt` file so LLMs can index your documentation efficiently, serves Markdown instead of HTML to AI agents, and exposes a standards-based API catalog for automated discovery. These features reduce token consumption and help agents process your content faster. +Your site is automatically optimized for AI tools and search engines. Fern hosts an `llms.txt` file, serves Markdown instead of HTML to AI agents, and exposes a standards-based API catalog, so agents index and process your content with fewer tokens. + +These files are generated from your content without calling a model. Your [`robots.txt`](/learn/docs/seo/robots-txt) controls whether third-party tools can crawl or train on them. @@ -51,3 +53,20 @@ Your site is automatically optimized for AI tools and search engines. Fern hosts + +## Data and controls + +Only the content a feature needs reaches a model: + +- [Ask Fern](/learn/docs/ai-features/ask-fern/overview#privacy-and-data-handling) sends the conversation and the documentation passages the viewer is allowed to read. Fern stores queries for analytics. +- [AI-generated examples](/learn/docs/ai-features/ai-examples) send the relevant part of your API definition. +- [Fern Agent](/learn/docs/ai-features/fern-agent#privacy-and-data-handling) sends your request and keeps it only during the session. + +Fern doesn't use your documentation content, your code, or your end users' queries to train AI models. + +| Feature | How to turn it off | +| --- | --- | +| Ask Fern, including the Slack app and MCP server | Select [**Disable** on the Ask AI card](/learn/docs/ai-features/ask-fern/overview#disable-ask-fern) in **Settings** in the [dashboard](https://dashboard.buildwithfern.com/). | +| MCP server only | Set [`page-actions.options.mcp`](/learn/docs/configuration/site-level-settings#page-actionsoptions.mcp) to `false` in `docs.yml`. | +| AI-generated examples | Set [`ai-examples.enabled`](/learn/docs/ai-features/ai-examples#disable-ai-generated-examples) to `false` in `docs.yml`. Hand-written examples always take precedence. | +| Third-party AI crawling of your published site | Configure [`robots.txt`](/learn/docs/seo/robots-txt). | diff --git a/fern/products/docs/pages/ask-fern/what-is-ask-fern.mdx b/fern/products/docs/pages/ask-fern/what-is-ask-fern.mdx index f9a1196a2..12be79ac1 100644 --- a/fern/products/docs/pages/ask-fern/what-is-ask-fern.mdx +++ b/fern/products/docs/pages/ask-fern/what-is-ask-fern.mdx @@ -119,6 +119,10 @@ Ask Fern can [mask structured personally identifiable information](/learn/docs/c +## Privacy and data handling + +Each request sends the conversation, the documentation passages retrieved for the question, the site domain, and any system prompt you configure to the model. Retrieval respects the viewer's [roles](/learn/docs/authentication/features/rbac), so answers draw only from content that viewer can read. Fern stores end-user queries and answers, including questions asked in Slack, to power analytics, and retains them indefinitely unless you set a [query log retention](#features) window. Fern doesn't use your documentation content or your end users' queries to train AI models. + ## Disable Ask Fern To turn Ask Fern off, open the [Fern Dashboard](https://dashboard.buildwithfern.com/), navigate to the **Settings** tab, and click **Disable** on the Ask AI card. This removes AI search and the side panel from your site.