Developers fix hidden barriers blocking SigNoz's built-in LLM observability feature

During a hackathon held July 20–26, 2026, a team discovered that SigNoz v0.133.0 already ships a complete LLM observability product in its open-source codebase, but several silent defaults prevent self-hosters from ever seeing it work. Two feature flags — enable_ai_observability and use_dashboard_v2 — are disabled by default, causing the UI to silently redirect or render blank dashboards even when the backend returns a success response. Additionally, two data registries that power model pricing and vendor span-mapping ship with zero rows, meaning dashboards display nothing out of the box. The team also identified a documentation mismatch where SigNoz's own CrewAI guide directs users to an instrumentation library that emits no compatible telemetry attributes. Their fix, enabling both flags together and seeding the empty registries, was upstreamed, and they also added answer-quality measurement as a previously missing observability signal.