Chatbot (Assistants API)
Use this component to integrate OpenAI's Assistants API into your Unity project. This allows AI agents to run tools, maintain threads, and perform multi-step actions through OpenAI’s hosted memory.
What it does
This component enables:
Text input & response using the OpenAI Assistants API
Voice input via Speech-to-Text (optional)
Voice output via Text-to-Speech (optional)
Image generation when requested by the assistant (optional)
Tool/function calling through OpenAI’s built-in tool invocation system
Multi-step reasoning using threads, runs, and memory management provided by the Assistants API
1. Assistant Setup
This section defines the identity and behavior of your assistant.
Selected Assistant
Reference to an Assistant object. Defines the assistant's configuration and tools.
Assistant Name
Display name of this assistant in the Editor. For UI purposes.
Description
Optional description shown in the Editor or UI.
Instructions
System prompt to define the assistant’s personality and role.
Tools
List of tools (function names) the assistant can use. Set in the Assistant object.
2. Assistant Behavior
These options control how the assistant responds.
Model
OpenAI model to use (e.g., GPT-4o). Set per run.
Response Format
Select the format of the assistant's response: text, json, json_schema, or auto. Use text for natural language replies, and json or json_schema for structured responses. auto lets the model choose the most suitable format.
Stream
Enable to receive streaming responses from the assistant.
3. Advanced Options
Optional fine-tuning of model behavior.
Temperature
Controls randomness. Higher = more creative.
Top P
Controls diversity. Lower = more focused.
4. Modules
Optional components that enhance the assistant’s capabilities.
Function Manager
Executes Unity functions when the assistant calls a tool.
Speech-to-Text
Enables voice input.
Text-to-Speech
Converts assistant responses to voice.
Image Generator
Allows assistant to generate images when appropriate.
5. Event Receivers
Use these to trigger UnityEvents in response to assistant activity.
Chat Event Receiver
Triggered when a message is sent or received.
Tool Call Receiver
Triggered when the assistant requests a tool call.
Streaming Event Receiver
Triggered as streamed tokens arrive.
Error Receiver
Triggered on any exception during chat.
6. Life Cycle Event Receivers
Receive callbacks for lower-level events in the Assistants API flow.
Assistant Event Receiver
Called on assistant load/update.
Thread Event Receiver
Called on thread creation/update.
Run Event Receiver
Called when a new run is started or completed.
Message Event Receiver
Called on message creation/receipt.
7. Required Action Handling
If the assistant uses required_actions (e.g., waits for a tool to complete), this section manages how Unity should respond.
Ignore Required Actions
If enabled, assistant won’t wait for function results.
Required Action Timeout
Time (in seconds) before a required action times out.
On Required Action
A list of UnityEvents to trigger when a required action is detected (e.g., tool call).
Last updated