Whisperer is a framework for building and orchestrating multi-agent AI systems or any form of workflow-related system.
At MonitorLizard (an opentelemetry-based Agentic AI-powered Observability platform), we needed a simple way of building and orchestrating observability-based agents, which help teams investigate and fix production issues. This led to the creation of Whisperer, which we’ve recently open-sourced.
At its core, Whisperer is a set of behaviours for defining agents and a fault-tolerant orchestrator capable of sequencing and executing these agents by leveraging the power of LLMs. Whisperer is a framework instead of a library; it lets you decide what implementations to use for your agents and sequencing strategy while abstracting away the hassle of orchestration.
In the ever-changing landscape of LLMs and Agentic systems, where implementation changes often, Whisperer serves as a solid blueprint for utilising improvements in the space while keeping the core stable.
Key Takeaways:
- Building and orchestrating multi-agent workflows
- Understanding AI agents and making multi-agent systems fault-tolerant
- Mental models for thinking about agents as functions or services.
Target Audience:
- Developers
- Academics
- Companies
- AI Enthusiasts