fix: add Amazonbot and MistralAI-Training to the AI training opt-out

docs/content/4.api/1.config.md

@@ -146,14 +146,15 @@ Some user-triggered agents may ignore robots.txt rules.
 
 - `false`: Adds no AI blocking group.
 - `true`: Blocks every token in `AiBots`, including search and user-directed retrieval agents.
-- `'training'`: Blocks `GPTBot`, `ClaudeBot`, `Google-Extended`, `meta-externalagent`, `Applebot-Extended`, and `CCBot`.
+- `'training'`: Blocks `GPTBot`, `ClaudeBot`, `Google-Extended`, `meta-externalagent`, `Amazonbot`, `MistralAI-Training`, `Applebot-Extended`, and `CCBot`.
 
 The `'training'` mode leaves dedicated search and answer agents unblocked, including OpenAI, Claude, and Perplexity agents.
 Other robots.txt groups and site indexing rules still apply.
 The module maintains this curated list locally, with operator sources in [`src/const.ts`](https://github.com/nuxt-modules/robots/blob/main/src/const.ts).
 
 [Google-Extended](https://developers.google.com/crawling/docs/crawlers-fetchers/google-common-crawlers#google-extended) controls both Gemini training and grounding.
 Blocking it also opts out of grounding in Gemini Apps and Vertex AI, while preserving Google Search inclusion.
+[Amazonbot](https://developer.amazon.com/amazonbot) may be used to train Amazon AI models. Amazon's search and answer agents, Amzn-SearchBot and Amzn-User, stay unblocked.
 [CCBot](https://commoncrawl.org/about) supplies a shared web archive used for training and other purposes.
 Robots.txt cannot guarantee that content stays out of all training datasets.
 

src/const.ts

@@ -74,6 +74,7 @@ export const AiBots = [
   'ImagesiftBot',
   'MistralAI-User',
   'MistralAI-Index',
+  'MistralAI-Training',
   'AI2Bot',
   'DuckAssistBot',
   'FirecrawlAgent',
@@ -96,6 +97,11 @@ export const AiTrainingBots = [
   'Google-Extended',
   // https://developers.facebook.com/docs/sharing/webmasters/web-crawlers/
   'meta-externalagent',
+  // https://developer.amazon.com/amazonbot
+  // Amazon search and answers use Amzn-SearchBot and Amzn-User, which Amazon says are not used for training.
+  'Amazonbot',
+  // https://docs.mistral.ai/robots
+  'MistralAI-Training',
   // https://support.apple.com/en-us/119829
   'Applebot-Extended',
   // https://commoncrawl.org/about

test/e2e/ai-training.test.ts

@@ -27,6 +27,8 @@ describe('training opt-out', () => {
       'Google-Extended',
       'Applebot-Extended',
       'meta-externalagent',
+      'Amazonbot',
+      'MistralAI-Training',
       'CCBot',
     ]))
     for (const bot of [

test/unit/const.test.ts

@@ -1,5 +1,5 @@
 import { describe, expect, it } from 'vitest'
-import { AiBots } from '../../src/const'
+import { AiBots, AiTrainingBots } from '../../src/const'
 
 describe('aiBots', () => {
   it('contains unique user-agent tokens', () => {
@@ -21,3 +21,18 @@ describe('aiBots', () => {
     expect(AiBots).not.toContain('GoogleOther')
   })
 })
+
+describe('aiTrainingBots', () => {
+  it('is a subset of AiBots, so `true` blocks everything `training` blocks', () => {
+    expect(AiBots).toEqual(expect.arrayContaining(AiTrainingBots))
+  })
+
+  it('includes the documented training crawlers', () => {
+    expect(AiTrainingBots).toEqual(expect.arrayContaining(['Amazonbot', 'MistralAI-Training']))
+  })
+
+  it('leaves dedicated search and user-directed agents out', () => {
+    for (const bot of ['Amzn-SearchBot', 'Amzn-User', 'MistralAI-Index', 'MistralAI-User'])
+      expect(AiTrainingBots).not.toContain(bot)
+  })
+})