ARCHITECTURE-1165: Insert Token. Get Wisdom. See the Trace - Observing your AI workloads

Insert token, ask a question, and watch the magic unfold. But behind every prediction, embedding lookup, and rendered image, a live meter is silently ticking. A looping agent or unexpected bottleneck can burn through your token budget in seconds. Ready to peek behind the curtain? Meet Malachite: a multi-agent fortune teller that processes vision, reasoning, embeddings, and image generation in real time.

Built on secure SUSE Application Collection and NVIDIA container images, Malachite is managed by SUSE Rancher Prime, running on SUSE Virtualization, and deployed with SUSE AI Factory for NVIDIA. We bring this workload to life while observed with SUSE Observability to track performance, vLLM metrics, and topology seamlessly.

By integrating OpenTelemetry, deep tracing unlocks operational intelligence across every agent, giving you and your CFO crystal-clear insights into token timing, health status, and cost per hop. Get wisdom, see the trace, and master your AI stack live on stage!