Core Components

Every agentic system consists of a few fundamental building blocks.

1. The Agent (The Brain)

The core logic that orchestrates the workflow. It holds the “system prompt” or persona and manages the context window.

2. Tools (The Hands)

Capabilities exposed to the agent. These can be:

• APIs: Weather, Stock Prices, Internal Databases.
• Functions: Calculator, Code Interpreter.
• Retrievers: RAG search against vector databases.

3. Memory (The Context)

• Short-term Memory: The current conversation history and scratchpad of thoughts.
• Long-term Memory: Vector databases or persistent storage for recalling past interactions.

Modeling in Sruja

We can map these components to Sruja elements:

• Agent -> container or component
• Tool -> component or external system
• Memory -> datastore

import { * } from 'sruja.ai/stdlib'


AgentSystem = system "Customer Support Bot" {
Brain = container "Orchestrator" {
  description "Main control loop"
}

Memory = container "Context Store" {
  ShortTerm = component "Conversation History"
  LongTerm = component "Vector DB"
}

Tools = container "Toolbelt" {
  CRM = component "CRM Connector"
  KB = component "Knowledge Base"
}

Brain -> Tools.CRM "Look up user"
Brain -> Tools.KB "Search policy"
Brain -> Memory.ShortTerm "Read/Write context"
}

view index {
include *
}