Uber has a combined interface for doing debugging, visualizing, and evals on an agent.
(This would be amazing to have at work)
Post
Uber has a combined interface for doing debugging, visualizing, and evals on an agent.
(This would be amazing to have at work)
Team is trying to provide the best tools for working in AI.
Their verticals:
- Agent Builder for rapid prototyping
- Write production-ready AI applications with AgentFX
- Monitor and deploy with Agent Studio
AgentFX is a standard interface to agents at Uber.
the runtime uses the OSS Agent Protocol (with langchain.ai)
Provides Threads, Runs, Messages, and Agents under one protocol.
Why use the Agent Protocol? Standardized:
- Chat history management
- Low Level APIs
- Observability (would love to know more about this)
Uber has a combined interface for doing debugging, visualizing, and evals on an agent.
(This would be amazing to have at work)
This tooling supports local servers, online servers, and versioning for rollbacks and testing.
Evals can even be run offline, or in CI/CD
A space for Bonfire maintainers and contributors to communicate