Traced Llm Proxy

A debugging and observability layer for Gemini calls via Vertex AI. This wraps each completion request in an OpenTelemetry trace span and returns both the LLM response and telemetry data, so you can instrument your AI workflows without changing how you call the model. Part of the GOSCE agent portfolio running on getvda.ai, it's designed for production deployments where you need visibility into latency, token usage, and request patterns. The server exposes the traced-gemini capability over streamable HTTP transport. Reach for this when you're running Gemini in production and want APM-grade traces without writing custom instrumentation for every LLM call.