Which observability bridge should AI Gateway ship first?

[MitulShah1]

AI Gateway has an OpenTelemetry foundation. The next step is bridge plugins for LLMOps and APM tools. We want to prioritize the integrations users actually need first.

Candidates

• LangSmith
• Langfuse
• Phoenix
• Helicone
• Datadog
• New Relic
• Sentry
• Honeycomb
• Grafana / Tempo

What we want feedback on

• Is OTLP enough for your use case, or do you need a full SDK-based bridge?
• Which fields are mandatory: trace ID, provider, model, cost, latency, cache hit, routing decision?
• What privacy level should be the default?
• Should bridge plugins send events, spans, logs, or all three?

flowchart LR
  App[App or framework] --> Gateway[AI Gateway]
  Gateway --> OTel[OTel spans]
  Gateway --> Events[Gateway events]
  OTel --> Collector[OTLP collector]
  Events --> Bridge[Bridge plugin]
  Bridge --> LangSmith
  Bridge --> Langfuse
  Bridge --> Phoenix
  Collector --> Grafana
  Collector --> Datadog

Loading

Please do not share private traces, credentials, or customer-specific telemetry in this thread.