SPIKE: Use of Lang Graph for the Agent Framework

[gideonaina]

Crew AI is currently being used for the agent orchestration framework. However, Crew AI is optimized for managing complex workflows, such as chaining tasks across different agents with distinct capabilities (e.g., summarizing, translating, analyzing).
However, for workflows that require clear dependencies, conditional task execution, and non-linear flows, such as research assistance or decision trees, Lang graph (https://langchain-ai.github.io/langgraph/) might be more suited. Seethe technical comparison below:

Technical Comparison

Feature	Crew AI	LangGraph
Workflow Structure	Sequential	Graph-based
Collaboration Between Agents	Strong focus	Moderate focus
Dependency Management	Limited	Advanced
Parallel Execution	Limited	Native
Debugging and Visualization	Basic	Robust (via graph structure)
Scalability	Excellent for linear flows	Excellent for complex flows

Since Lang graph supports more complex workflow than Crew AI, the goal of this spike is to come up with a design to that will use Lang graph for the agentic framework.