As enterprises scale Generative AI into production, operations teams face a major hurdle: traditional APM tools only monitor server uptime, leaving IT blind to LLM-specific performance issues, sudden token cost spikes, and silent failures.
In this session, we will showcase how DX Operational Observability (DX O2) delivers native GenAI and LLM Observability (up to release 26.7.1) to close the visibility gap between core IT infrastructure and modern AI applications. We will cover:
- GenAI stack Layers: How to capture telemetry across the Orchestration (LangChain), Semantic (ChromaDB), and Model (Ollama, OpenAI) tiers.
- Four Operational Mandates: A practical look at tracking LLM token economics, optimizing RAG pipeline triage, monitoring Model Reliability (including Time to First Token), and mapping unified topologies.
- Demo : A first look at how DX O2 detects and traces complex live incidents—including vector database outages, RAG retrieval bottlenecks, and malicious prompt injection alerts.
Speakers:
- Ashish Aggarwal, Product Lead (Ingestion & Infra Observability)
- Srinivas Venkata Bevara, Technical Lead (Ingestion)