What problem does it solve?
LLM application traces often arrive broken or incomplete: empty inputs and outputs, spans disconnected from parent traces, and missing labels or user identifiers. This Skill provides a systematic workflow to inspect production traces with the LangWatch CLI, identify the root cause of instrumentation gaps, and verify fixes.
Core Features & Use Cases
- Trace Inspection: Fetch recent traces with
langwatch trace search and drill into individual traces with langwatch trace get to check inputs, outputs, spans, and metadata.
- Issue Diagnosis: Detect common problems such as empty input/output, disconnected spans, missing labels, and absent user_id or thread_id, with a cause-to-fix mapping table.
- Guided Fixes and Verification: Compare your setup against framework-specific integration docs via
langwatch docs, apply fixes, and re-inspect traces to confirm improvement.
- Use Case: Your Python app sends traces to LangWatch but every trace shows
<empty> input/output. Use this Skill to confirm the symptom, identify the missing autotrack_openai_calls(client) configuration, apply the fix, and verify new traces are complete.
Quick Start
Ask the agent to inspect my recent LangWatch traces from the last week and diagnose why inputs and outputs are empty.
