Some documentation for span_attributes.

Author: sstanovnik

docs/logfire.md

@@ -356,3 +356,38 @@ Agent.instrument_all(instrumentation_settings)
 ```
 
 This setting is particularly useful in production environments where compliance requirements or data sensitivity concerns make it necessary to limit what content is sent to your observability platform.
+
+### Adding custom attributes to agent spans
+
+If you want to add metadata to and agent span, pass the [`span_attributes`][pydantic_ai.models.instrumented.InstrumentationSettings] argument.
+You can either provide a static `dict[str, str]`:
+
+```python {title="instrumentation_span_attributes_static.py"}
+from pydantic_ai import Agent
+from pydantic_ai.models.instrumented import InstrumentationSettings
+
+agent = Agent(
+    'openai:gpt-5',
+    instrument=InstrumentationSettings(
+        span_attributes={
+            'gen_ai.agent.description': 'A fiery agent!',
+            'my_org.tenant.id': '42',
+        }
+    )
+)
+```
+
+Alternatively, supply a callable that receives the run [`RunContext`][pydantic_ai.tools.RunContext] so attributes can be derived from the prompt, dependencies, or usage data collected during the run:
+
+```python {title="instrumentation_span_attributes_callable.py"}
+from pydantic_ai import Agent, RunContext
+from pydantic_ai.models.instrumented import InstrumentationSettings
+
+def span_attribute_callback(ctx: RunContext[None]) -> dict[str, str]:
+    return {'my_org.prompt_preview': str(ctx.prompt)[:100]}
+
+agent = Agent(
+    'openai:gpt-5-nano',
+    instrument=InstrumentationSettings(span_attributes=span_attribute_callback)
+)
+```