How persona interpretation works
This article explains how CustomGPT.ai automatically converts your agent's persona into focused instructions and how those instructions improve agent performance.
This article explains how CustomGPT.ai automatically converts your agent's persona into focused instructions, and how those instructions improve agent performance.
What persona interpretation does
Every time you save your agent's persona, the platform analyzes the text and generates sub-personas in the background. Sub-personas are focused sets of instructions derived from your main persona - each targeting a specific aspect of how the agent thinks and operates.
What sub-personas are generated
Two sub-personas are currently generated from your persona:
- Persona notes - A summary of the key characteristics and intent in your persona.
- Request routing map - Instructions for which tool or action the agent should use in different situations.
How sub-personas are used
When your agent processes a message, it uses several internal processes - each responsible for a different part of handling the conversation, such as planning how to answer, generating the response, or deciding which action to take. These are referred to as sub-agents.
Instead of passing your full persona text to every sub-agent, each sub-agent receives only the sub-personas relevant to its role. This makes instructions more precise and targeted.
Expert mode
Expert mode lets power users view and manually edit sub-personas. It is not intended for everyday use.
Warning: Editing sub-personas incorrectly can cause your agent to behave in unexpected ways. Most builders will never need to use Expert mode. If you are not certain what you are changing and why, do not use it.
See Edit sub-personas using Expert mode for details.
Related articles
Updated about 6 hours ago