Customizing Agent Knowledge & Behaviour
Make Sapience Agents your own
Customizing Sapience Knowledge & Behavior
Sapience has a powerful memory system, covering 4 Tiers and treats facts and preferences as separate. This gives much better real-world results than the way Chat GPT does it.
All Agents are built around knowledge & behavior. This is covered extensively in the Agents overview articles. What we are focussed on here is setting global rules for both behavior, as well as fact-based data for Sapience to inject into all conversations (its invisible to you).
To get started with this, click on Menu > Settings & Customization. You can then help control both the behavior of all Agents, as well as what it knows about you.
Behavior: the first thing you can tweak is how you want all Sapience Agents to behave. So long as it doesn’t run counter to their core programming, all Agents in the system will follow your guidelines. Telling AI models how you want them to behave (or “steering”) has a surprisingly powerful impact on their behavior at runtime. So, put in here a few instructions that you want to apply to all chats. Whilst you can put in “always answer in poetry and rhyming verse”, you’re better off using something like the 1st screenshot below - but make it your own.
User context: Getting good answers out of AI systems is all about giving the model the right context. We do this for you in many places throughout the app, but here is one place that you should take the time to tell Sapience about yourself. Use who, what, when, why, where if you need a framework. But tell Sapience who you are, what you do, what your job is, what your priorities are at the moment. Armed with that information, every chat you have will be far more on-point, and tailored to what you currently care about.
To go deeper into our Memory architecture, including managing stored memories and the 4 Layers available, see the dedicated article on Memory:
You can also customize the Project Agent separately, and that’s covered in this article: