[AI-01] Implement RAG retrieval and AI agent orchestration

[Krisor1986]

Goal

Create grounded AI responses with traceable sources and human approval.

Tasks

• Load agent configuration and instructions
• Retrieve relevant knowledge chunks
• Build the model context
• Generate answers with source references
• Enforce human approval for sensitive actions
• Add handoff and audit logs

Acceptance criteria

• Answers use only authorized company knowledge
• Sources can be traced
• Low-confidence or sensitive cases can reach a human

Dependencies

• [KNOW-01]
• [INBOX-01]