Equip EVI with Tools to enable function calling during Chats.

EVI empowers your application by enabling the use of external functions—referred to as Tools—to enhance conversations. Tools allow EVI to trigger custom logic (for example, searching the web, fetching weather data, or updating a database) based on user input.

Tool use is only supported when specifying certain supplemental LLMs within your configuration. Currently, our function calling feature supports Anthropic and OpenAI models.

For best results, we suggest choosing a fast and intelligent LLM that performs well on function calling benchmarks. On account of its speed and intelligence, we recommend Claude 3.5 Haiku as the supplemental LLM in your EVI configuration when using Tools. Function calling is not available if you are using your own custom language model. We plan to support more function calling LLMs in the future.

See our API reference for details on how to configure Tools and Built-in Tools in your EVI configuration.

Tool types

  • Function Tools: These are custom functions you create and manage via our API. With Function Tools, you are responsible for defining the function, specifying its input parameters (using a JSON schema), and invoking it when EVI signals that a function call is needed.
  • Built-in Tools: These Tools are natively integrated into EVI and do not require you to define or invoke any function logic. Hume provides built-in Tools such as:
    • web_search: Searches the web for real-time information when needed.
    • hang_up: Closes the WebSocket connection with a status code of 1000 (normal closure), typically triggered when the conversation has ended.

Creating tools

While we recommend using our Platform UI for creating Tools, you can also create and manage Function Tools programmatically through our API. For complete details on creating a Tool via the API, see the API reference.

Implementing tools

For a step-by-step walkthrough on creating Tools and understanding how both Function Tools and Built-in Tools are implemented within an EVI Chat session, refer to our Tool Use Guide.


Built with