chore: import upstream snapshot with attribution

2026-07-13 13:34:48 +08:00
commit 77bb5bf71f
3762 changed files with 353249 additions and 0 deletions
@@ -0,0 +1,45 @@
+import { type Client } from '@botpress/client'
+import { z } from '@bpinternal/zui'
+import { Tool } from 'llmz'
+
+/** Extract the full content & the metadata of the specified pages as markdown. */
+export const browsePages = (client: Client) =>
+  new Tool({
+    name: 'browser_browsePages',
+    description: 'Extract the full content & the metadata of the specified pages as markdown.',
+    input: z
+      .object({
+        urls: z.array(z.string()),
+        waitFor: z
+          .default(
+            z
+              .number()
+              .describe(
+                'Time to wait before extracting the content (in milliseconds). Set this value higher for dynamic pages.'
+              ),
+            350
+          )
+          .describe(
+            'Time to wait before extracting the content (in milliseconds). Set this value higher for dynamic pages.'
+          ),
+      })
+      .catchall(z.never()),
+    output: z
+      .object({
+        results: z.array(
+          z
+            .object({
+              url: z.string(),
+              content: z.string(),
+              favicon: z.optional(z.string()),
+              title: z.optional(z.string()),
+              description: z.optional(z.string()),
+            })
+            .catchall(z.never())
+        ),
+      })
+      .catchall(z.never()),
+    handler: async (input) => {
+      return client.callAction({ type: 'browser:browsePages', input }).then(({ output }) => output) as any
+    },
+  })
@@ -0,0 +1,23 @@
+import { type Client } from '@botpress/client'
+import { z } from '@bpinternal/zui'
+import { Tool } from 'llmz'
+
+/** Capture a screenshot of the specified page. */
+export const captureScreenshot = (client: Client) =>
+  new Tool({
+    name: 'browser_captureScreenshot',
+    description: 'Capture a screenshot of the specified page.',
+    input: z
+      .object({
+        url: z.string(),
+      })
+      .catchall(z.never()),
+    output: z
+      .object({
+        imageUrl: z.string(),
+      })
+      .catchall(z.never()),
+    handler: async (input) => {
+      return client.callAction({ type: 'browser:captureScreenshot', input }).then(({ output }) => output) as any
+    },
+  })
@@ -0,0 +1,3 @@
+export * from './browsePages'
+export * from './captureScreenshot'
+export * from './webSearch'
@@ -0,0 +1,102 @@
+import { type Client } from '@botpress/client'
+import { z } from '@bpinternal/zui'
+import { Tool } from 'llmz'
+
+/** Search information on the web. You need to browse to that page to get the full content of the page. */
+export const webSearch = (client: Client) =>
+  new Tool({
+    name: 'browser_webSearch',
+    description: 'Search information on the web. You need to browse to that page to get the full content of the page.',
+    input: z
+      .object({
+        query: z.string().min(1, undefined).max(1000, undefined).describe('What are we searching for?'),
+        includeSites: z
+          .optional(
+            z
+              .array(
+                z
+                  .string()
+                  .regex(/^(?:[a-z0-9](?:[a-z0-9-]{0,61}[a-z0-9])?\.)+[a-z]{2,}$/, undefined)
+                  .min(3, undefined)
+                  .max(50, undefined)
+              )
+              .max(20, undefined)
+              .describe('Include only these domains in the search (max 20)')
+          )
+          .describe('Include only these domains in the search (max 20)'),
+        excludeSites: z
+          .optional(
+            z
+              .array(
+                z
+                  .string()
+                  .regex(/^(?:[a-z0-9](?:[a-z0-9-]{0,61}[a-z0-9])?\.)+[a-z]{2,}$/, undefined)
+                  .min(3, undefined)
+                  .max(50, undefined)
+              )
+              .max(20, undefined)
+              .describe('Exclude these domains from the search (max 20)')
+          )
+          .describe('Exclude these domains from the search (max 20)'),
+        count: z
+          .default(
+            z
+              .number()
+              .min(1, undefined)
+              .max(20, undefined)
+              .describe('Number of search results to return (default: 10)'),
+            10
+          )
+          .describe('Number of search results to return (default: 10)'),
+        freshness: z
+          .optional(z.enum(['Day', 'Week', 'Month']).describe('Only consider results from the last day, week or month'))
+          .describe('Only consider results from the last day, week or month'),
+        browsePages: z
+
+          .boolean()
+          // .default(false)
+          .describe('Whether to browse to the pages to get the full content'),
+      })
+      .catchall(z.never()),
+    output: z
+      .object({
+        results: z.array(
+          z
+            .object({
+              name: z.string().describe('Title of the page'),
+              url: z.string().describe('URL of the page'),
+              snippet: z.string().describe('A short summary of the page'),
+              links: z
+                .optional(
+                  z
+                    .array(
+                      z
+                        .object({
+                          name: z.string(),
+                          url: z.string(),
+                        })
+                        .catchall(z.never())
+                    )
+                    .describe('Useful links on the page')
+                )
+                .describe('Useful links on the page'),
+              page: z.optional(
+                z
+                  .object({
+                    url: z.string(),
+                    content: z.string(),
+                    favicon: z.optional(z.string()),
+                    title: z.optional(z.string()),
+                    description: z.optional(z.string()),
+                  })
+                  .catchall(z.never())
+              ),
+            })
+            .catchall(z.never())
+        ),
+      })
+      .catchall(z.never()),
+    handler: async (input) => {
+      return client.callAction({ type: 'browser:webSearch', input }).then(({ output }) => output) as any
+    },
+  })
@@ -0,0 +1,193 @@
+import { Client } from '@botpress/client'
+import { z } from '@bpinternal/zui'
+import { ObjectInstance, Tool } from 'llmz'
+
+const sanitizePath = (path: string) => {
+  // Remove leading/trailing slashes and replace multiple slashes with a single slash
+  path = path.replace(/(^\/+|\/+$)/g, '').replace(/\/{2,}/g, '/')
+  // Allow only alphanumeric, underscore, hyphen, and slashes
+  path = path.replace(/[^a-zA-Z0-9_\-\\/\\.]/g, '')
+  // Ensure the path does not start with a slash
+  path = path.startsWith('/') ? path : `/${path}`
+  // Ensure the path does not end with a slash
+  path = path.endsWith('/') ? path.slice(0, -1) : path
+  // Ensure the path does not contain consecutive slashes
+  path = path.replace(/\/{2,}/g, '/')
+  // Ensure the path does not contain a leading dot
+  path = path.replace(/^\.\//, '')
+  // Ensure the path does not contain a trailing dot
+  path = path.replace(/\/\.$/, '')
+
+  return path.toLowerCase().trim()
+}
+
+const getPathFolder = (path: string) => {
+  const parts = path.split('/')
+  parts.pop() // Remove the last part (file name)
+  return parts.join('/') || '/'
+}
+
+const parseFilePath = (path: string) => {
+  if (!path || typeof path !== 'string') {
+    throw new Error(`Invalid file path: "${path}". Path must be a non-empty string.`)
+  }
+
+  if (path.length > 255) {
+    throw new Error(`Invalid file path: "${path}". Path length exceeds 255 characters.`)
+  }
+
+  const sanitizedPath = sanitizePath(path)
+
+  if (!sanitizedPath.length) {
+    throw new Error(`Invalid file path: "${path}". Path cannot be empty.`)
+  }
+
+  const folder = getPathFolder(sanitizedPath)
+  const fileName = sanitizedPath.split('/').pop()?.trim() || ''
+
+  if (!fileName.length) {
+    throw new Error(`Invalid file path: "${path}". File name cannot be empty.`)
+  }
+
+  if (fileName.length && !fileName.includes('.')) {
+    throw new Error(`Invalid file path: "${fileName}". File name must contain an extension.`)
+  }
+
+  return {
+    path: sanitizedPath,
+    folder,
+    file: fileName,
+  }
+}
+
+export const makeFileSystem = (client: Client) =>
+  new ObjectInstance({
+    name: 'fs',
+    description: 'File system operations',
+    tools: [
+      new Tool({
+        name: 'readFile',
+        description: 'Read a file from the file system',
+        input: z.string().describe('File path to read'),
+        output: z.string().describe('File content'),
+        handler: async (path) => {
+          const { path: key } = parseFilePath(path)
+
+          const { file } = await client.getFile({
+            id: key,
+          })
+
+          return await fetch(file.url).then((res) => res.text())
+        },
+      }),
+
+      new Tool({
+        name: 'writeFile',
+        description: 'Write a file to the file system',
+        input: z.object({
+          path: z.string().describe('File path to write'),
+          content: z.string().describe('File content to write'),
+        }),
+        handler: async (input) => {
+          const { path, file, folder } = parseFilePath(input.path)
+
+          await client.uploadFile({
+            key: path,
+            content: input.content,
+            accessPolicies: ['public_content'],
+            publicContentImmediatelyAccessible: true,
+            tags: {
+              purpose: 'file-system',
+              folder: folder || '/',
+              name: file,
+            },
+          })
+        },
+      }),
+
+      new Tool({
+        name: 'listFiles',
+        description: 'Write a file to the file system',
+        input: z.string().describe('Folder path to list files from').default('/'),
+        output: z
+          .array(
+            z.object({
+              folder: z.string().describe('Folder path'),
+              file: z.string().describe('File name'),
+              createdAt: z.string().describe('File creation date'),
+              updatedAt: z.string().describe('File last update date'),
+              size: z.number().describe('File size in bytes'),
+              url: z.string().url().describe('File URL'),
+              contentType: z.string().describe('File content type'),
+            })
+          )
+          .describe('List of files in the folder'),
+        handler: async (path) => {
+          const folder = sanitizePath(path || '/')
+
+          const files = await client.list
+            .files({
+              sortDirection: 'desc',
+              sortField: 'updatedAt',
+              tags: {
+                purpose: 'file-system',
+                folder,
+              },
+            })
+            .collect({ limit: 1000 })
+
+          return files.map((file) => ({
+            folder: file.tags?.folder || '/',
+            file: file.tags?.name || file.key,
+            createdAt: file.createdAt,
+            updatedAt: file.updatedAt,
+            size: file.size,
+            url: file.url,
+            contentType: file.contentType,
+          }))
+        },
+      }),
+
+      new Tool({
+        name: 'deleteFile',
+        description: 'Delete a file from the file system',
+        input: z.string().describe('File path to delete'),
+        output: z.boolean().describe('True if file was deleted successfully'),
+        handler: async (path) => {
+          const { path: key } = parseFilePath(path)
+          await client.deleteFile({ id: key })
+          return true
+        },
+      }),
+
+      new Tool({
+        name: 'moveFile',
+        description: 'Rename a file in the file system',
+        input: z.object({
+          before: z.string().describe('Current file path'),
+          after: z.string().describe('New file path'),
+        }),
+        handler: async ({ before, after }) => {
+          const { path: oldKey } = parseFilePath(before)
+          const { path: newKey, file, folder } = parseFilePath(after)
+
+          await client.copyFile({
+            idOrKey: oldKey,
+            destinationKey: newKey,
+            overwrite: true,
+          })
+
+          await client.deleteFile({ id: oldKey })
+
+          await client.updateFileMetadata({
+            id: newKey,
+            tags: {
+              purpose: 'file-system',
+              folder: folder || '/',
+              name: file,
+            },
+          })
+        },
+      }),
+    ],
+  })