AI agent context overview
AI agent's context is the information you provide in addition to the AI service's model. This additional context is key to defining the agent's purpose, actions and constraints. The AI agent's context consists of two layers of information sent to the AI service provider: The system context and additional context provided in the Run an agent module:
- System context: This is the main definition of how the agent works, what is its purpose and what are its constraints. The system context contains:
- The information from the AI agent's system prompt in the agent settings.
- Names and descriptions of the system provided to the AI agent in the Agent settings.
- Descriptions of the system inputs and outputs.
- Additional context: This is additional information sent to the agent by the specific Run an agent module. The additional context contains:
- Additional system instructions from the module settings.
- Names and descriptions of the provided to the AI agent in the module settings.
- Descriptions of the agent's system inputs and outputs.
- Messages you send to the AI agent.
The system and additional context layers form a definition hierarchy. The system context is in the agent's settings and is always a part of the agent's information. You can customize the default, system context by providing more information in the additional context.
For example, an agent that manages a shop inventory might have always available a to list the shop inventory. You would provide the agent with access to this as a system . In where you need to just check the inventory the agent could have access only to this .
In a dedicated to ordering new items the agent could get access to another for creating orders.