Skip to main content
ThinkEx provides a comprehensive set of AI tools that enable the assistant to interact with your workspace, search the web, process files, and more.

​
Tool Factory

Tools are created using the createChatTools factory function: 
import { createChatTools } from "@/lib/ai/tools";

const tools = createChatTools({
  workspaceId: "uuid",
  userId: "uuid",
  activeFolderId: "uuid",
  threadId: "uuid",
  clientTools: {}, // Frontend tools
  enableDeepResearch: false,
});

​
Workspace Tools

​
createNote

Create a new note card in the workspace.
​
title
string
required
The title of the note card
​
content
string
required
The markdown body content. DO NOT repeat title in content. Start with subheadings/text.
​
sources
array
Optional sources from web search or deep research
Example Usage: 
await createNote({
  title: "Photosynthesis Overview",
  content: "## Process\n\nPhotosynthesis converts light energy...",
  sources: [{
    title: "Biology Textbook",
    url: "https://example.com/biology",
  }]
});

​
updateNote

Update an existing note with targeted edits or full rewrite.
​
noteName
string
required
The name of the note to update (fuzzy matched)
​
oldString
string
required
Text to find. Use '' for full rewrite. For targeted edit: exact text from readWorkspace Content section.
​
newString
string
required
Replacement text (entire note if oldString is empty)
​
replaceAll
boolean
default:false
Replace every occurrence of oldString; use for renaming or changing repeated text
​
title
string
New title for the note. If not provided, existing title is preserved.
Example Usage: 
// Full rewrite
await updateNote({
  noteName: "My Note",
  oldString: "",
  newString: "# New Content\n\nCompletely new note body..."
});

// Targeted edit
await updateNote({
  noteName: "My Note",
  oldString: "old section content",
  newString: "updated section content"
});

​
deleteItem

Permanently delete a card/note from the workspace by name.
​
itemName
string
required
Item name or virtual path (e.g., pdfs/Report.pdf) to delete

​
selectCards

Select cards by their titles and add them to conversation context.
​
cardTitles
array
required
Array of card titles to search for and select
Response:
​
success
boolean
Whether the operation succeeded
​
addedCount
number
Number of cards successfully selected
​
message
string
Descriptive message about the selection result

​
Flashcard Tools

​
createFlashcards

Create a new flashcard deck.
​
title
string
The title of the flashcard deck (defaults to ‘Flashcard Deck’ if not provided)
​
cards
array
required
Array of flashcard objects, each with ‘front’ and ‘back’ properties (minimum 1 card)
Example Usage: 
await createFlashcards({
  title: "Spanish Vocabulary",
  cards: [
    { front: "Hello", back: "Hola" },
    { front: "Goodbye", back: "Adiós" },
    { front: "Thank you", back: "Gracias" },
  ]
});

​
updateFlashcards

Add more flashcards to an existing flashcard deck and/or update its title.
​
deckName
string
required
The name or ID of the flashcard deck to update
​
cards
array
Array of flashcard objects to add
​
title
string
New title for the flashcard deck

​
Quiz Tools

​
createQuiz

Create an interactive quiz. Extract topic from user message. Use selected cards as context if available.
​
topic
string
The topic or specific focus instructions for the quiz - extract from user’s message
​
contextContent
string
Content from selected cards in system context if available
​
questionCount
number
default:5
The specific number of questions requested by the user (1-50)
​
sourceCardIds
array
IDs of source cards
​
sourceCardNames
array
Names of source cards
Example Usage: 
await createQuiz({
  topic: "Linear Algebra Basics",
  questionCount: 10,
});

​
updateQuiz

Update quiz title and/or add more questions. Can use new topic, selected cards, or general knowledge.
​
quizName
string
required
The name of the quiz to update (fuzzy matched)
​
title
string
New title for the quiz
​
topic
string
New topic for questions
​
contextContent
string
Content from newly selected cards in system context
​
questionCount
number
Number of questions to add (1-50)

​
Search & Code Tools

​
webSearch

Search the web for current information using Google Search with grounding.
​
query
string
required
The search query to look up on the web (1-500 characters)
Response:
​
text
string
Synthesized search results with summary and key findings
​
sources
array
Array of source objects with title and url
​
groundingMetadata
object
Google grounding metadata with chunks
Example Usage: 
const result = await webSearch({
  query: "latest developments in quantum computing 2024"
});

​
executeCode

Execute Python code for calculations, data processing, algorithms, or mathematical computations.
​
task
string
required
Description of the task to solve with code
Example Usage: 
await executeCode({
  task: "Calculate the fibonacci sequence up to n=20"
});

​
searchWorkspace

Grep-like text search across workspace content (notes, flashcards, PDFs, quizzes, audio transcripts).
​
pattern
string
required
Search pattern (plain text or regex) — the term/phrase to find in workspace content
​
include
string
Filter by item type, e.g., "note", "flashcard", "pdf"
​
path
string
Virtual path prefix to scope search, e.g., "Physics/" for items under Physics folder
Example Usage: 
await searchWorkspace({
  pattern: "photosynthesis",
  include: "note",
  path: "Biology/"
});

​
readWorkspace

Read content of a workspace item (note, flashcard deck, PDF summary, quiz) by path or name.
​
path
string
Virtual path (e.g., Physics/notes/Thermodynamics.md) — unambiguous when duplicates exist
​
itemName
string
Name for fuzzy match — use when path unknown
​
lineStart
number
default:1
1-based line number to start from. Use with limit for pagination.
​
limit
number
default:500
Max lines to return (max 2000). Use with lineStart for pagination.
​
pageStart
number
For PDFs only: 1-indexed start page (e.g., 5 for page 5)
​
pageEnd
number
For PDFs only: 1-indexed end page inclusive (e.g., 10 for pages 5–10)
Example Usage: 
// Read note
await readWorkspace({
  itemName: "My Study Notes"
});

// Read specific PDF pages
await readWorkspace({
  path: "pdfs/Textbook.pdf",
  pageStart: 10,
  pageEnd: 15
});

​
File & URL Processing Tools

​
processFiles

Process and analyze files including PDFs, images, documents, and videos.
​
urls
array
Array of file/video URLs to process (Supabase storage URLs or YouTube URLs)
​
fileNames
array
Workspace item names or virtual paths (e.g., "Annual Report" or "pdfs/Annual Report.pdf")
​
pdfImageRefs
array
Images from PDFs — map placeholder names to base64 for analysis
​
forceReprocess
boolean
default:false
Set to true to bypass cached PDF content and re-analyze the file
Supported File Types:
  • PDFs: Extracted via Azure Document AI OCR (cached after first extraction)
  • Images: JPEG, PNG, GIF, WebP, SVG
  • Videos: YouTube URLs (native Gemini support)
  • Documents: DOC, DOCX, TXT
Example Usage: 
// Process workspace files
await processFiles({
  fileNames: ["Research Paper", "Lecture Slides"]
});

// Process external URLs
await processFiles({
  urls: [
    "https://storage.supabase.co/...",
    "https://youtube.com/watch?v=..."
  ]
});

// Extract image from PDF
await processFiles({
  pdfImageRefs: [{
    pdfName: "Syllabus",
    imageId: "img-0.jpeg"
  }]
});

​
processUrls

Analyze web pages using Google’s URL Context API.
​
jsonInput
string
required
JSON string containing an object with ‘urls’ (array of web URLs).Example: '{"urls": ["https://example.com"]}'
Features:
  • Extracts content from regular web URLs (http/https)
  • Supports up to 20 URLs per request
  • Automatically falls back between scraping methods
  • Returns analyzed content with sources
Example Usage: 
await processUrls({
  jsonInput: JSON.stringify({
    urls: [
      "https://en.wikipedia.org/wiki/Machine_learning",
      "https://blog.example.com/ai-trends"
    ]
  })
});

​
YouTube Tools

​
searchYoutube

Search for YouTube videos.
​
query
string
required
The search query for YouTube videos
Response:
​
success
boolean
Whether the search succeeded
​
videos
array
Array of video objects with id, title, thumbnail, etc.

​
addYoutubeVideo

Add a YouTube video to the workspace.
​
videoId
string
required
The YouTube Video ID (not the full URL)
​
title
string
required
The title of the video
Example Usage: 
await addYoutubeVideo({
  videoId: "dQw4w9WgXcQ",
  title: "Introduction to Neural Networks"
});

​
Tool Context

All workspace tools receive a shared context: 
export interface WorkspaceToolContext {
  workspaceId: string | null;    // Current workspace UUID
  userId: string | null;          // Authenticated user UUID
  activeFolderId?: string;        // Active folder UUID
  threadId?: string | null;       // Conversation thread UUID
}
This context provides:
  • Security: Ensures user has access to workspace
  • Scoping: Operations limited to current workspace
  • Context: Tools can reference active folder for new items

​
Error Handling

All tools return consistent error structures: 
{
  success: false,
  message: "Descriptive error message",
  // Additional context fields
}
Common error scenarios:
  • Item not found: Suggests available items or paths
  • No workspace context: User must be in a workspace
  • Permission denied: User lacks access to workspace
  • Invalid input: Validation error with requirements

Build docs developers (and LLMs) love

Get started for freeTalk to us