In Adventure 03: The AI Observatory, you made it to RaviHyral aboard the Perihelion (ART). But now Outpost Verada’s monitoring team is complaining — ART’s traces are a mess. Non-standard span names, missing token usage, and Jaeger drowning in noise from every single request.
Your mission? Fix the instrumentation, record errors properly, and filter out the noise.
Expert: The Noise Filter
In this level, you will take ART’s observability from “technically working” to closer to production-ready.
What You’ll Learn:
- GenAI Semantic Conventions: How to follow the OpenTelemetry GenAI spec for LLM spans, including token usage attributes.
- Tail Sampling: How to configure the OpenTelemetry Collector to keep only meaningful traces (errors and slow requests in this case) and drop the rest.
The Tech Stack:
- Python & Ollama (Local LLM)
- LangChain & Qdrant (Vector Database)
- OpenTelemetry & OpenLLMetry
- Jaeger
- Kubernetes (pre-provisioned)
New Verification System: Verify your solution instantly inside your Codespace and get your Certificate of
Completion!
Ready to Start?
Good luck, SecUnit. ART is judging your span names. 
Credits: The characters of this adventure are borrowed from the
fantastic Murderbot Diaries series by Martha Wells! 
