Function Calling

Function Calling allows NovaAI models to interact with external tools, APIs, and services. Instead of only generating text, the model can decide when to invoke a function, pass structured arguments, and use the returned data to generate a final response.

Overview

Function Calling extends the capabilities of language models by connecting them to real-world actions. For example, an agent can check the weather, retrieve database records, send emails, or perform calculations by calling predefined functions.

Important: Models do not execute functions directly. They generate a structured function call, and your application is responsible for executing it and returning the result.

Function Calling Flow

Rendering diagram…

Example Function

1{
2  "name": "get_weather",
3  "parameters": {
4    "city": "San Francisco"
5  }
6}

Your application receives this object, executes the appropriate function, and sends the result back to the model.

Supported Tool Types

Tool	Description
API Calls	Retrieve data from external services.
Database	Query application data.
File System	Read or write files.
Custom Functions	Execute your own business logic.

Example Use Cases

Weather assistants
Calendar scheduling
Database lookups
Order tracking
CRM integrations
AI workflow automation

Keep function names descriptive and return structured JSON. Predictable inputs and outputs make tool execution more reliable.

Before Enabling Function Calling

Define clear function names.
Validate all input parameters.
Return consistent JSON responses.
Handle failures gracefully.
Restrict access to trusted tools only.

Next Step

Continue to Fine Tuning to learn how to customize NovaAI models for your specific datasets and use cases.

Function Calling

Overview

Function Calling Flow

Example Function

Supported Tool Types

Example Use Cases

Before Enabling Function Calling

Next Step

Related Documentaion

Agents

Streaming

Chat Completion