Skip to content
Last updated

Function Calling

🔨 In Development — This section is still being developed and may change.
Freddy assistants can invoke external tools and business logic through structured function calls. Use this guide to define functions, describe parameters, and safely handle tool results.

When to use function calling

  • Combine model reasoning with deterministic operations (databases, APIs, workflows).
  • Enforce schema validation for user inputs before executing sensitive actions.
  • Provide assistants with controlled access to enterprise systems.

Implementation steps

  1. Register tools in Freddy Hub or via the API with clear names and JSON schemas.
  2. Supply function definitions when creating responses or assistants (functions array).
  3. Inspect the model’s tool call payload, execute your backend logic, and return formatted results.
  4. Resume the conversation with the tool output so Freddy can craft the final response.
 
Â