Package ai.tabforge.workshop.orchestrator

package ai.tabforge.workshop.orchestrator
Orchestration engine — the agentic loop that coordinates all sub-agents.

Classes in this package implement the complete review lifecycle:

   INIT → DECOMPOSE → DISPATCH → COLLECT → EVALUATE → ESCALATE → AGGREGATE → COMPLETE
  • OrchestratorAgent — drives the loop; owns session state
  • TaskDecomposer — routes files to the right specialist
  • ContextWindowManager — enforces token budgets per agent call
  • AgentResultAggregator — merges + deduplicates findings across agents

CERTIFICATION NOTE — Agentic Architecture & Orchestration (27% of exam): OrchestratorAgent is the central exam artifact for this domain. The lifecycle stages above map directly to the agentic loop patterns described in the certification course.

CERTIFICATION NOTE — Context Management & Reliability (15% of exam): ContextWindowManager and OrchestratorAgent#escalate() together implement the two key reliability patterns: token budget management and human-in-the-loop escalation.

Implemented in Phase 1 Day 7-10.

  • Classes
    Class
    Description
    AgentResultAggregator
    Merges findings from all sub-agents into a single, deduplicated list.
    ContextWindowManager
    Enforces token budget constraints for each sub-agent API call.
    OrchestratorAgent
    The central coordinator of the Workshop review — manages all sub-agents, enforces workflow, handles lifecycle hooks, and decides when to escalate.
    ReviewSession
    When Claude AI periodically calls GetReportTool in order to obtain review status (and so, when GetReportTool calls
    invalid reference 
    OrchestratorAgent#getStatus()
    ) the GetReportToll will return JSON filled with data from this class.
    TaskDecomposer