Agent Core

The Overseer agent is the brain of the platform — an autonomous AI system that can plan, reason, and execute multi-step tasks.

How It Works

Overseer uses Vercel AI SDK with a tool loop pattern:

1. User sends a message (via Telegram, Discord, Web, or API)
2. Agent receives the message with conversation context
3. Agent decides whether to respond directly or use tools
4. If tools are needed, the agent executes them and processes results
5. The agent may loop multiple times for complex, multi-step tasks
6. Final response is streamed back to the user

Tool Loop

Unlike simple chatbots, Overseer uses agentic tool loops where the AI can:

• Execute multiple tools in sequence
• Use tool outputs to inform next decisions
• Retry or adjust approach on failures
• Chain complex workflows automatically

// Simplified agent flow
const result = await generateText({
  model: provider(modelId),
  system: soulPrompt,
  messages: conversationHistory,
  tools: availableTools,
  maxSteps: 10, // Allow up to 10 tool calls per turn
});

Personality (SOUL.md)

The agent's behavior is defined by SOUL.md — a Markdown file that acts as the system prompt:

• Expertise — What the agent is good at
• Personality — Communication style and tone
• Guardrails — Things the agent should never do
• Preferences — Default behaviors and approaches

See SOUL.md guide for details.