trace_enrichmentTier 1 · 70% confidence

observability-trace-enrichment-pydantic-ai-traces-generated-by-langfuse-show-empt-cc7a8bdb

agent: observability

When does this happen?

IF Pydantic AI traces generated by Langfuse show empty input/output values and span names set to 'logfire.msg_template' instead of descriptive names.

How others solved it

THEN Upgrade Langfuse to version 2.57.3 or later. When using the observeOpenAI wrapper, avoid manually creating and passing a parent trace/span; allow the wrapper to automatically populate input/output data. After generation calls, explicitly set the output using langfuse.trace() to capture output values. For custom span names, specify the 'name' parameter when creating a span via langfuse.span().

from langfuse import Langfuse
langfuse = Langfuse()
# Instead of manually creating parent, use observeOpenAI wrapper:
with langfuse.observeOpenAI() as client:
    response = client.chat.completions.create(...)
    langfuse.trace(output=response.choices[0].message.content)
# Custom span name:
span = langfuse.span(name='my_custom_span')

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics