packages/sample-app/sample_app/agno_streaming_example.py (1)

7-7: Consider moving initialization inside the main block.

Traceloop.init() at module level runs on import, which can cause issues when the module is imported for testing or other purposes. Consider moving it inside the if __name__ == "__main__" block for better isolation.

-Traceloop.init(app_name="agno_streaming_example")
-
-
 def get_weather(location: str) -> str:

Then in the main block:

if __name__ == "__main__":
    Traceloop.init(app_name="agno_streaming_example")
    asyncio.run(test_streaming())

packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/streaming.py (2)

19-37: Unused constructor parameters.

instance is stored in _self_instance (line 31) and token_histogram is stored in _self_token_histogram (line 34), but neither is used anywhere in the class. Consider removing these if they're not needed, or document their intended future use.

     def __init__(
         self,
         span,
         response,
-        instance,
         start_time,
         duration_histogram: Histogram = None,
-        token_histogram: Histogram = None,
     ):
         super().__init__(response)

         self._self_span = span
-        self._self_instance = instance
         self._self_start_time = start_time
         self._self_duration_histogram = duration_histogram
-        self._self_token_histogram = token_histogram
         self._self_events = []
         self._self_final_result = None
         self._self_instrumentation_completed = False

Note: If these parameters are intentionally kept for future use, please add a comment indicating so.

---

119-219: Consider extracting shared instrumentation logic to reduce duplication.

AgnoStream._complete_instrumentation is nearly identical to AgnoAsyncStream._complete_instrumentation. Consider extracting the shared logic into a helper function or mixin to improve maintainability.

Example helper approach:

def _finalize_instrumentation(
    span, final_result, start_time, duration_histogram, 
    instrumentation_completed_flag_setter
):
    """Shared instrumentation completion logic."""
    # Common logic here...

This would reduce the duplication between lines 65-116 and 168-219.

packages/opentelemetry-instrumentation-agno/tests/cassettes/test_agent/test_agent_arun_streaming.yaml (1)

96-100: Consider scrubbing session cookies from cassettes.

The Set-Cookie headers contain Cloudflare session tokens (__cf_bm, _cfuvid). While these are likely ephemeral and not directly exploitable, per coding guidelines for VCR cassettes, consider scrubbing or replacing these with placeholder values to follow best practices for not committing potentially sensitive data.

Verify that the VCR recording configuration includes appropriate filters for sensitive headers. You might configure VCR to filter these automatically:

@vcr.use_cassette(..., filter_headers=['set-cookie', 'authorization'])

packages/opentelemetry-instrumentation-agno/tests/cassettes/test_agent/test_agent_run_streaming.yaml (1)

96-100: Consider scrubbing session cookies from cassettes.

Same observation as the async streaming cassette - the Set-Cookie headers contain Cloudflare session tokens. Consider configuring VCR to filter these headers automatically for consistency with security best practices.

packages/opentelemetry-instrumentation-agno/tests/test_agent.py (2)

174-175: Consider adding a defensive check before indexing.

The list comprehension [s for s in spans if "agent" in s.name][-1] will raise IndexError if no agent spans are found, making test failures harder to diagnose.

-    agent_span = [s for s in spans if "agent" in s.name][-1]
+    agent_spans = [s for s in spans if "agent" in s.name]
+    assert agent_spans, "Expected at least one agent span"
+    agent_span = agent_spans[-1]

---

239-240: Same IndexError risk as the sync variant.

Apply the same defensive check pattern here for consistency.

-    agent_span = [s for s in spans if "agent" in s.name][-1]
+    agent_spans = [s for s in spans if "agent" in s.name]
+    assert agent_spans, "Expected at least one agent span"
+    agent_span = agent_spans[-1]

packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/__init__.py (2)

185-187: Move import time to module level.

The time import is repeated inside function bodies at lines 185, 234, 334, and 384. While Python caches imports, placing it at module level is cleaner and avoids repeated dictionary lookups.

Add at module level with other imports:

import time

Then remove the local imports inside the methods.

---

163-183: Consider extracting common span attribute setup.

The attribute-setting logic for GEN_AI_SYSTEM, TRACELOOP_SPAN_KIND, agent name, model, and input message is duplicated between streaming (lines 163-183) and non-streaming (lines 213-232) branches. The same pattern repeats in _AgentARunWrapper. A helper method could reduce this duplication.

def _set_agent_span_attributes(span, instance, args):
    """Set common span attributes for agent runs."""
    span.set_attribute(GenAIAttributes.GEN_AI_SYSTEM, "agno")
    span.set_attribute(
        SpanAttributes.TRACELOOP_SPAN_KIND,
        TraceloopSpanKindValues.AGENT.value,
    )
    if hasattr(instance, "name"):
        span.set_attribute(GenAIAttributes.GEN_AI_AGENT_NAME, instance.name)
    if hasattr(instance, "model") and instance.model:
        model_name = getattr(
            instance.model, "id", getattr(instance.model, "name", "unknown")
        )
        span.set_attribute(GenAIAttributes.GEN_AI_REQUEST_MODEL, model_name)
    if args and should_send_prompts():
        span.set_attribute(SpanAttributes.TRACELOOP_ENTITY_INPUT, str(args[0]))

Also applies to: 213-232

Configuration used: CodeRabbit UI

Review profile: CHILL

Plan: Pro

Disabled knowledge base sources:

• Linear integration is disabled by default for public repositories

You can enable these sources in your CodeRabbit configuration.