ai-orchestrator

name	ai-orchestrator
description	Expert implementation of AI workflows using LangGraph and CrewAI. Use when building multi-agent systems, stateful graphs, or complex tool-use scenarios.

AI Orchestrator (LangGraph & CrewAI)

LangGraph Workflow

• Stateful Graphs: Define a clear State TypedDict.
• Nodes: Each node should be a single, testable function that returns state updates.
• Edges: Use conditional edges for routing based on LLM output or tool results.
• Persistence: Use MemorySaver for thread-level state persistence.

CrewAI Orchestration

• Role-Based Agents: Give agents distinct personas and goals.
• Tasks: Define granular tasks with clear expected outputs.
• Process Types: Use Process.sequential for linear workflows and Process.hierarchical for manager-led structures.

Best Practices

• Tool Use: Wrap all external interactions (searches, file edits) in LangChain tools.
• Error Handling: Implement retry logic for LLM calls and tool failures within the graph nodes.