GenAI Observability for Splunk

GenAI Observability for Splunk provides production monitoring for LLM, RAG, and agent applications — track token usage, latency, cost, retrieval quality, and errors across OpenAI, Anthropic, LangChain, and more. **Works with any OpenTelemetry GenAI pipeline — no custom SDK required.** If you already emit OpenTelemetry GenAI traces (gen_ai.* semantic conventions), point your OpenTelemetry Collector at Splunk and the dashboards light up — with no changes to your application code. Prefer a turnkey setup? An optional lightweight Python SDK instruments your code with simple decorators. Key Features: - Ingests standard OpenTelemetry GenAI traces (gen_ai.* semantic conventions) via the OpenTelemetry Collector - Real-time monitoring of LLM calls, RAG pipelines, and AI agents - Token usage tracking with cost estimation (bundled model-pricing lookup) - Latency analysis with P50, P95, P99 percentiles and tokens/sec - Error detection and alerting - Pre-built dashboards for immediate insights - Optional Python SDK for one-line, decorator-based instrumentation - Works with OpenAI, Anthropic, LangChain, LlamaIndex, and any OpenTelemetry-instrumented framework - Data model acceleration to speed up reporting and dashboard performance on extremely large datasets Use Cases: - Monitor production AI applications - Optimize token usage and reduce costs - Debug slow or failing LLM calls - Track AI spend by model, workflow, or user - Ensure SLA compliance for AI services Bring your existing OpenTelemetry pipeline, or use the optional SDK — either way you get ready-to-use dashboards, model-pricing lookups, and a collector recipe to get started in minutes. Docs, examples, and the collector configuration: https://github.com/rootiq-ai/splunk_genai_observability/