Description
Trace the behaviour of a tool-using language-model agent to investigate failures that ordinary health checks may miss. The course instruments search, retrieval, calculation and summarization steps with OpenTelemetry, then compares Phoenix and Langfuse tracing backends deployed through Docker. Spans, events, attributes and context propagation provide a record of how a request moves through the system.
The debugging exercises address empty search results, hidden retries, stale information and incorrect answers that do not produce an explicit exception. Other lessons connect evaluations with traces, examine prompt-injection signals and break down latency and cost by workflow segment. Incident exercises, sampling decisions, ownership and review practices extend the technical work into an observability process that a team can maintain.
Intended audience: AI and software engineers debugging deployed LLM applications. Infrastructure engineers and evaluation developers investigating agent traffic.
What you will learn
- Instrument an agent's tool calls and context flow with OpenTelemetry.
- Compare traces from Docker-hosted Phoenix and Langfuse backends.
- Investigate silent failures and connect response evaluations to execution traces.
- Analyze latency, usage cost and sampling decisions at different workflow levels.
- Document incidents and establish team practices for reviewing agent behaviour.
Course specifications
- Publisher
- Udemy
- Instructors
- Frank Robotics Lab
- Level
- Beginner to Advanced
- Duration
- 7 hours and 35 minutes
- Lecture count
- 71
- Subtitle
- None
- Quality
- 1080p
- Edition
- 2026-9
Course topics
View course topics
The course topics image is temporarily unavailable.
Prerequisites
- Comfort with Python functions, classes, dictionaries and tracebacks.
- An API key for the OpenAI-compatible model interface used by the exercises.
- Docker and Docker Compose for the tracing backends.
- Previous OpenTelemetry experience is not required.
Pictures
This picture is temporarily unavailable.
Introduction video
How to open the files
Keep all parts of the same edition together before extracting the first archive. Open the included course media with a compatible player.
Download resources
Choose one edition and keep its parts together. Alternate providers are available when listed.
2026-9 · 1080p · 1
Direct downloads
2026-9 · 1080p
2026-9 · 1080p · 2
Rapidgator
2026-9 · 1080p
Size
3.9 GB
Leave a Reply
Signed-in member comments appear immediately.