How tools extend your AI assistant

Tools Overview

Tools let your AI chatbot reach out to external databases or perform real-world actions during a chat conversation. They transform your chatbot into a powerful business agent that can solve real customer problems.

Read-Only vs. Write-Action Tools

Vindex supports two categories of tools based on what they do:

1. Read-Only Tools (Information Gathering)

These tools fetch real-time information to help the chatbot answer customer questions accurately.
  • Examples: Fetching a product price, checking an order shipping status, or searching store hours.
  • Uses
    text
    GET
    or
    text
    POST
    requests and caches responses (
    text
    cache_ttl
    ) for fast performance.

2. Write-Action Tools (Task Execution)

These tools perform changes or tasks in external software.
  • Examples: Creating a support ticket, booking an appointment, or subscribing a customer to a newsletter.
  • Safety: Can be configured to require confirmation from the user or admin before running.
  • Uses
    text
    POST
    ,
    text
    PUT
    ,
    text
    PATCH
    , or
    text
    DELETE
    requests.

Powerful Built-in Upgrades

Every Vindex chatbot automatically gets these premium tools to process numbers, display charts, and search the live web:

1. Built-in Math Engine (
text
compute_logic
)

Allows the AI to perform complex mathematical calculations and aggregate numbers on the spot without making mistakes.
  • Example: Customer asks: "What was my usage growth between March and April?" The chatbot computes sums and growth percentages instantly.

2. Interactive Charts & Tables (
text
render_ui_component
)

Allows the chatbot to present visual, high-fidelity components inside the chat box instead of boring blocks of text.
  • Capability: Can render interactive Bar Charts, Pie Charts, Data Tables, and Stat Cards directly to visitors.

3. Internet Web Search (
text
search_the_web
)

Allows your chatbot to step outside its static files and search the live internet to find current news, live stock prices, or competitive research.

Human-in-the-Loop Confirmation

For high-stakes actions (such as reserving a booking, changing grades, or initiating payments), you can toggle Requires Confirmation:
  1. The chatbot recognizes it needs to use a tool (e.g., booking a room).
  2. The chatbot pauses and shows the details to the customer: "I'm about to book Room 302 for tomorrow at 2:00 PM. Please confirm if that's correct."
  3. The action is only fired once the customer clicks Confirm.

Custom Software & Database Integrations

Do you want your chatbot to speak to your custom billing system, school management portal, clinic software, or internal CRM?
Your developer can easily configure Custom Tools to connect the AI chatbot securely. These tools are authorized using encrypted passwords (headers) so that your private API keys remain 100% hidden and safe from visitors.
To see custom headers, REST/GraphQL protocols, and dynamic request parameters, toggle Developer Mode at the top of the sidebar.

Custom API Tools (Developer Only)

You can define custom tools for any Channel. Each tool supports:
  • Protocols: REST, GraphQL, or Form POST.
  • Dynamic Parameters: Define query parameters or request bodies that the AI will extract dynamically from the text.
  • Response Mapping: Use
    text
    response_path
    (dot notation, e.g.
    text
    data.tracking.status
    ) to extract exactly what the AI needs from complex JSON responses.

Secure Integrations via Custom Headers

You can authorize Vindex to speak to your internal systems securely by utilizing custom HTTP Headers:
  1. Authorization: Set headers such as
    text
    Authorization: Bearer your-secret-token
    in the tool configuration card.
  2. Encrypted at Rest: Headers are stored encrypted using AES-256 on our backend. The client chatting with the AI never sees these keys; they are injected by the Gateway only during the outgoing request to your API.

Example: School Management Interface

  • Read Tool: A teacher asks, "Who is John's guardian?" The chatbot fires a
    text
    GET
    request using the authorized headers, retrieves the information, and replies.
  • Write Tool: The teacher says, "Set attendance: John is Present." The chatbot prepares the API payload. Because this changes records, the Gateway halts the request, presents a confirmation dialogue, and only fires the write action after the teacher confirms.

Limits & Best Practices

  • Keep Responses Small: Keep tool outputs under 3,000 characters to prevent blowing the AI's prompt size budget.
  • Tool Descriptions: Write clear, plain-English descriptions (e.g. "Use this tool to find shipping details for a specific tracker ID"). The AI reads this description to decide when to call the tool.
  • snake_case Names: Use lowercase snake_case for tool names (e.g.
    text
    check_order_status
    ).
PreviousSelf-LearningNextAction Tools