Publisher
Arize AI
Product Details
This agent brings Arize AX, an observability and evaluation platform for LLM applications and agents, into Gemini. Engineering and product teams can ask what their AI systems are doing, find what is broken, and get to root cause without leaving Google Workspace, with answers grounded in real production traces, spans, and sessions from their Arize AX projects.
Built on the open Agent-to-Agent (A2A) protocol, Alyx works as a governed peer to Gemini, and every request runs under the signed-in user’s own Arize permissions. Engineers can ask for error spans from the last 7 days or traces with latency over 1000ms, filter traces by content, metadata, status, latency, or cost, and pull the full span tree for a trace to see where it failed, stalled, or overspent. Results appear as structured tables and summaries. Admins get user-delegated access through OAuth 2.0 against the Arize OIDC server, per-user RBAC enforcement, a standard A2A agent card with JSON-RPC transport, and no migration.
Key Use Cases
LLM Application Latency and Bottleneck Diagnostics
AI engineers analyze Arize AX execution spans directly within Gemini to identify slow retrieval steps, model inference delays, and unresponsive agent tools.
Production Agent Error Tracing and Debugging
Development teams isolate tool execution failures and hallucination patterns across production agent sessions using structured trace-span trees and user RBAC controls.
Explore detailed deployment path
Requires Gemini. Access integration prerequisites, specialized agent configuration guides, and implementation documentation.