map-reduce-llm

Using Redis A2A, we perform deep research on topics, it recursively generates sub-tasks launching agents which research multiple sub-topics based on your main questions. We manage these Agents using redis-a2a task queues and tavily to perform web searches. The advantage of using task queues instead of langgraph / deep agents is that we can use other agents to monitor the task graph, dynamically adding subtasks / removing redundant tasks.

This is useful because deep research is not very interactive and not recursively updating its sources based on what it finds, while our recursive approach is both deeper and better informed than the approach deep research typically takes which is Find sources -> Research -> Summarize. Our approach keeps going back and forth between those steps, while another agents keeps monitoring progress helping you manage costs!