Adhere to new /llms.txt standard

docs/.vuepress/config.ts

@@ -8,6 +8,7 @@ import { registerComponentsPlugin } from '@vuepress/plugin-register-components';
 import { getDirname, path } from '@vuepress/utils';
 import { balancerTheme } from '../../theme/';
 import { navbar, sidebar } from './configs/index.js';
+import { llmsTxtPlugin } from './plugins/llmsTxtPlugin.js';
 
 const isProd = process.env.NODE_ENV === 'production';
 
@@ -72,6 +73,7 @@ export default defineUserConfig({
 
   // use plugins
   plugins: [
+    llmsTxtPlugin(),
     registerComponentsPlugin({
       componentsDir: path.resolve(__dirname, './components'),
     }),

docs/.vuepress/plugins/llmsTxtPlugin.ts

@@ -0,0 +1,107 @@
+import type { Plugin } from '@vuepress/core'
+import { fs, path } from '@vuepress/utils'
+import { globby } from 'globby'
+import { unified } from 'unified'
+import remarkParse from 'remark-parse'
+import remarkStringify from 'remark-stringify'
+import remarkFrontmatter from 'remark-frontmatter'
+import { visit } from 'unist-util-visit'
+import { toMarkdown } from 'mdast-util-to-markdown'
+import type { Heading } from 'mdast'
+import { directiveToMarkdown } from 'mdast-util-directive'
+import remarkDirective from 'remark-directive'
+import { gfmToMarkdown } from 'mdast-util-gfm'
+
+/**
+ * Generates the following files and serves them at the appropriate paths:
+ * - llms.txt
+ * - llms-full.txt
+ * - .md version of each docs page like https://docs.balancer.fi/concepts/core-concepts/introduction.md
+ */
+export const llmsTxtPlugin = (): Plugin => ({
+  name: 'llms-txt-generator',
+
+  onGenerated: async (app) => {
+
+    const outDir = app.dir.dest()
+    const docsDir = app.dir.source()
+
+    const glob = `${docsDir}/**/*.md`
+    const files = await globby(glob);
+
+    const content = [`# Balancer V3`, '', '> Learn, integrate, and build on a programmable AMM', '']
+    const llmsTxtContent = [...content, '## Docs', ''];
+    const llmsFullTxtContent = content;
+
+    for (const file of files) {
+      let pagePath = `${file.replace(docsDir, '').replace(/\.[^.]*$/, '')}.md`
+      if (!pagePath) continue
+
+
+      const contents = await fs.readFile(file, 'utf-8');
+      const parser = unified().use(remarkParse).use(remarkFrontmatter).use(remarkDirective).use(remarkStringify)
+
+      // Abstract syntax tree from the contents of all the markdown files
+      const ast = parser.parse(contents)
+
+      // For links to pages served at /llms.txt
+      visit(ast, { type: 'heading', depth: 1 }, (n, i) => {
+        const node = n.children[0]
+        if (node.type !== 'text') return
+
+        const value = node.value
+        const [, title, subTitle] = value.match(/^([^\[\]]+)(?: \[([^\[\]]+)\])?$/) ?? []
+
+        // Look for a description: either use subtitle or first paragraph after heading
+        let found = false
+        let description = subTitle
+        if (!description)
+          visit(ast, { type: 'paragraph' }, (n, j) => {
+            if (found) return                    
+            if (j && i && j <= i) return         
+
+            found = true
+            description = toMarkdown(n, {extensions: [directiveToMarkdown(), gfmToMarkdown()]})
+
+            // Remove container directive syntax ( only for llms.txt )
+            description = description.replace(/^.*\\:::.*(?:\n|$)/gm, '').replace(/\n+$/, '');
+          })
+
+        llmsTxtContent.push(`- [${title}](https://docs.balancer.fi${pagePath})${description ? `: ${description}` : ''}`)
+      })
+
+      // Adjust depth of headings to make sense given site title is H1
+      visit(
+        ast,
+        (n) => n.type === 'heading',
+        (n) => {
+          const node = n as Heading
+          if (node.depth === 1 || node.depth === 2 || node.depth === 3 || node.depth === 4)
+            node.depth = (node.depth + 1) as 2 | 3 | 4 | 5
+        },
+      )
+
+      // remove front matter
+      visit(ast, { type: 'yaml' }, (_, i, p) => {
+        if (!p) return
+        if (typeof i !== 'number') return
+        p.children.splice(i, 1)
+      })
+
+      const llmFriendlyMarkdown = toMarkdown(ast, {extensions: [directiveToMarkdown(), gfmToMarkdown()]})
+
+      // Serve the .md version of each page alongside the .html
+      const mdOutputPath = path.join(outDir, pagePath)
+      fs.writeFileSync(mdOutputPath, llmFriendlyMarkdown) 
+
+      llmsFullTxtContent.push( llmFriendlyMarkdown, '')
+    }
+
+    const llmsTxt = llmsTxtContent.join('\n')
+    const llmsFullTxt = llmsFullTxtContent.join('\n')
+
+    fs.ensureDirSync(outDir)
+    fs.writeFileSync(path.resolve(outDir, 'llms.txt'), llmsTxt)
+    fs.writeFileSync(path.resolve(outDir, 'llms-full.txt'), llmsFullTxt)
+  },
+}) 

docs/README.md

@@ -39,4 +39,4 @@ features:
     details: All things data
     link: /data-and-analytics/README.md
 
-
+---

docs/developer-reference/contracts/error-signatures.md

@@ -4,6 +4,8 @@ order: 4
 
 # Error signatures
 
+Catalogue for decoding custom error signatures into their associated error names
+
 | Error Name                           | Parameters                                                   | Error Signature |
 | ------------------------------------ | ------------------------------------------------------------ | --------------- |
 | AfterAddLiquidityHookFailed          |                                                              | `0x1f11c6e8`    |

docs/partner-onboarding/balancer-v2/point-programs.md

@@ -6,6 +6,7 @@ order: 7
 # Setting Up a Partner Points Program Category on the Balancer UI
 
 This guide will walk you through the process of setting up a category for your partner points program in the Balancer protocol. By following these steps, you'll be able to tag sets of pools in the Balancer API and frontend, allowing users to earn points for providing liquidity to specific pools.
+
 :::info
 All information provided herein is referencing how to make changes to the [metadata repository ](https://github.com/balancer/metadata/) to register a points program on the Balancer Zen UI.
 :::