Skip to content

fix(agno): add streaming support for Agent.run() and Agent.arun() #3483

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
nirga merged 1 commit into from
Dec 1, 2025
Merged

fix(agno): add streaming support for Agent.run() and Agent.arun() #3483

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
270 changes: 190 additions & 80 deletions packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@
_FunctionCallAExecuteWrapper,
)
from opentelemetry.instrumentation.agno.config import Config
from opentelemetry.instrumentation.agno.streaming import AgnoAsyncStream, AgnoStream
from opentelemetry.instrumentation.agno.utils import (
dont_throw,
should_send_prompts,
Expand Down Expand Up @@ -149,12 +150,16 @@ def __call__(self, wrapped, instance, args, kwargs):
) or context_api.get_value("suppress_agno_instrumentation"):
return wrapped(*args, **kwargs)

span_name = f"{getattr(instance, 'name', 'unknown')}.agent"
is_streaming = kwargs.get("stream", False)

if is_streaming:
span_name = f"{getattr(instance, 'name', 'unknown')}.agent"

span = self._tracer.start_span(
Copy link
Copy Markdown
Member

@doronkopit5 doronkopit5 Dec 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why not start_as_current_span?

Copy link
Copy Markdown
Member Author

@nirga nirga Dec 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@doronkopit5 bwcause it will be ended outside of this method (only when the stream ends)

span_name,
kind=SpanKind.CLIENT,
)
Comment on lines +155 to +161
Copy link
Copy Markdown

@coderabbitai coderabbitai Bot Dec 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🔴 Critical

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Check streaming.py for span/context handling
rg -nP --type=py "start_span|set_current|attach|context|use_span|Token" packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/streaming.py

Repository: traceloop/openllmetry

Length of output: 47

🏁 Script executed:

#!/bin/bash
# Check if streaming.py imports or uses any context-related OpenTelemetry modules
rg -nP --type=py "from.*context|import.*context" packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/streaming.py

Repository: traceloop/openllmetry

Length of output: 47

🏁 Script executed:

#!/bin/bash
# Look at the full streaming.py file to understand the structure
wc -l packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/streaming.py

Repository: traceloop/openllmetry

Length of output: 160

🏁 Script executed:

#!/bin/bash
# Check how tool spans are created within the streaming context
rg -nP --type=py "tool.*span|_create_tool|wrap_tool" packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/__init__.py -A 5 | head -50

Repository: traceloop/openllmetry

Length of output: 47

🏁 Script executed:

#!/bin/bash
# Read streaming.py to understand the implementation
cat -n packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/streaming.py

Repository: traceloop/openllmetry

Length of output: 9702

🏁 Script executed:

#!/bin/bash
# Also check the imports in __init__.py to understand span management
head -50 packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/__init__.py | cat -n

Repository: traceloop/openllmetry

Length of output: 2160

🏁 Script executed:

#!/bin/bash
# Look for tool execution in the streaming branch
sed -n '155,180p' packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/__init__.py | cat -n

Repository: traceloop/openllmetry

Length of output: 1272

🏁 Script executed:

#!/bin/bash
# Check tool wrapper implementation
head -80 packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/_tool_wrappers.py | cat -n

Repository: traceloop/openllmetry

Length of output: 4007

🏁 Script executed:

#!/bin/bash
# Look for start_span calls in tool wrappers
rg -n "start_span|get_span|context" packages/opentelemetry-instrumentation-agno/opentelemetry/instrumentation/agno/_tool_wrappers.py -A 3 | head -60

Repository: traceloop/openllmetry

Length of output: 970

Span context propagation is broken for tool calls in streaming mode.

The streaming branch creates the agent span with start_span (line 158) but never sets it as the current context. Tool wrappers (_tool_wrappers.py line 34) use start_as_current_span for tool spans, which relies on the current OpenTelemetry context to establish parent-child relationships. Since the agent span is not set as the current context, tool spans will not be correctly parented to the agent span.

The non-streaming branch uses start_as_current_span (line 208), which properly establishes context. To fix this in streaming mode, attach the agent span to the context before wrapping the response. Use get_current_span() and token-based context attachment or wrap the stream iteration with the span context.


with self._tracer.start_as_current_span(
span_name,
kind=SpanKind.CLIENT,
) as span:
try:
span.set_attribute(GenAIAttributes.GEN_AI_SYSTEM, "agno")
span.set_attribute(
Expand All @@ -181,51 +186,100 @@ def __call__(self, wrapped, instance, args, kwargs):

start_time = time.time()

result = wrapped(*args, **kwargs)
response = wrapped(*args, **kwargs)

duration = time.time() - start_time
return AgnoStream(
span,
response,
instance,
start_time,
self._duration_histogram,
self._token_histogram,
)

if hasattr(result, "content") and should_send_prompts():
except Exception as e:
span.set_status(Status(StatusCode.ERROR, str(e)))
span.record_exception(e)
span.end()
raise
else:
span_name = f"{getattr(instance, 'name', 'unknown')}.agent"

with self._tracer.start_as_current_span(
span_name,
kind=SpanKind.CLIENT,
) as span:
try:
span.set_attribute(GenAIAttributes.GEN_AI_SYSTEM, "agno")
span.set_attribute(
SpanAttributes.TRACELOOP_ENTITY_OUTPUT, str(result.content)
SpanAttributes.TRACELOOP_SPAN_KIND,
TraceloopSpanKindValues.AGENT.value,
)

if hasattr(result, "run_id"):
span.set_attribute("agno.run.id", result.run_id)
if hasattr(instance, "name"):
span.set_attribute(GenAIAttributes.GEN_AI_AGENT_NAME, instance.name)

if hasattr(result, "metrics"):
metrics = result.metrics
if hasattr(metrics, "input_tokens"):
span.set_attribute(
GenAIAttributes.GEN_AI_USAGE_INPUT_TOKENS,
metrics.input_tokens,
if hasattr(instance, "model") and instance.model:
model_name = getattr(
instance.model, "id", getattr(instance.model, "name", "unknown")
)
if hasattr(metrics, "output_tokens"):
span.set_attribute(GenAIAttributes.GEN_AI_REQUEST_MODEL, model_name)

if args and should_send_prompts():
input_message = str(args[0])
span.set_attribute(
GenAIAttributes.GEN_AI_USAGE_OUTPUT_TOKENS,
metrics.output_tokens,
SpanAttributes.TRACELOOP_ENTITY_INPUT, input_message
)
if hasattr(metrics, "total_tokens"):

import time

start_time = time.time()

result = wrapped(*args, **kwargs)

duration = time.time() - start_time

if hasattr(result, "content") and should_send_prompts():
span.set_attribute(
SpanAttributes.LLM_USAGE_TOTAL_TOKENS, metrics.total_tokens
SpanAttributes.TRACELOOP_ENTITY_OUTPUT, str(result.content)
)

span.set_status(Status(StatusCode.OK))

self._duration_histogram.record(
duration,
attributes={
GenAIAttributes.GEN_AI_SYSTEM: "agno",
SpanAttributes.TRACELOOP_SPAN_KIND: TraceloopSpanKindValues.AGENT.value,
},
)
if hasattr(result, "run_id"):
span.set_attribute("agno.run.id", result.run_id)

if hasattr(result, "metrics"):
metrics = result.metrics
if hasattr(metrics, "input_tokens"):
span.set_attribute(
GenAIAttributes.GEN_AI_USAGE_INPUT_TOKENS,
metrics.input_tokens,
)
if hasattr(metrics, "output_tokens"):
span.set_attribute(
GenAIAttributes.GEN_AI_USAGE_OUTPUT_TOKENS,
metrics.output_tokens,
)
if hasattr(metrics, "total_tokens"):
span.set_attribute(
SpanAttributes.LLM_USAGE_TOTAL_TOKENS, metrics.total_tokens
)

span.set_status(Status(StatusCode.OK))

self._duration_histogram.record(
duration,
attributes={
GenAIAttributes.GEN_AI_SYSTEM: "agno",
SpanAttributes.TRACELOOP_SPAN_KIND: TraceloopSpanKindValues.AGENT.value,
},
)

return result
return result

except Exception as e:
span.set_status(Status(StatusCode.ERROR, str(e)))
span.record_exception(e)
raise
except Exception as e:
span.set_status(Status(StatusCode.ERROR, str(e)))
span.record_exception(e)
raise


class _AgentARunWrapper:
Expand All @@ -238,19 +292,23 @@ def __init__(self, tracer, duration_histogram, token_histogram):
self._token_histogram = token_histogram

@dont_throw
async def __call__(self, wrapped, instance, args, kwargs):
def __call__(self, wrapped, instance, args, kwargs):
"""Wrap the Agent.arun() call with tracing instrumentation."""
if context_api.get_value(
context_api._SUPPRESS_INSTRUMENTATION_KEY
) or context_api.get_value("suppress_agno_instrumentation"):
return await wrapped(*args, **kwargs)
return wrapped(*args, **kwargs)

span_name = f"{getattr(instance, 'name', 'unknown')}.agent"
is_streaming = kwargs.get("stream", False)

if is_streaming:
span_name = f"{getattr(instance, 'name', 'unknown')}.agent"

span = self._tracer.start_span(
Copy link
Copy Markdown
Member

@doronkopit5 doronkopit5 Dec 1, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

same question

span_name,
kind=SpanKind.CLIENT,
)

with self._tracer.start_as_current_span(
span_name,
kind=SpanKind.CLIENT,
) as span:
try:
span.set_attribute(GenAIAttributes.GEN_AI_SYSTEM, "agno")
span.set_attribute(
Expand All @@ -277,51 +335,103 @@ async def __call__(self, wrapped, instance, args, kwargs):

start_time = time.time()

result = await wrapped(*args, **kwargs)
response = wrapped(*args, **kwargs)

duration = time.time() - start_time

if hasattr(result, "content") and should_send_prompts():
span.set_attribute(
SpanAttributes.TRACELOOP_ENTITY_OUTPUT, str(result.content)
)

if hasattr(result, "run_id"):
span.set_attribute("agno.run.id", result.run_id)
return AgnoAsyncStream(
span,
response,
instance,
start_time,
self._duration_histogram,
self._token_histogram,
)

if hasattr(result, "metrics"):
metrics = result.metrics
if hasattr(metrics, "input_tokens"):
span.set_attribute(
GenAIAttributes.GEN_AI_USAGE_INPUT_TOKENS,
metrics.input_tokens,
)
if hasattr(metrics, "output_tokens"):
span.set_attribute(
GenAIAttributes.GEN_AI_USAGE_OUTPUT_TOKENS,
metrics.output_tokens,
)
if hasattr(metrics, "total_tokens"):
except Exception as e:
span.set_status(Status(StatusCode.ERROR, str(e)))
span.record_exception(e)
span.end()
raise
else:
async def async_wrapper():
span_name = f"{getattr(instance, 'name', 'unknown')}.agent"

with self._tracer.start_as_current_span(
span_name,
kind=SpanKind.CLIENT,
) as span:
try:
span.set_attribute(GenAIAttributes.GEN_AI_SYSTEM, "agno")
span.set_attribute(
SpanAttributes.LLM_USAGE_TOTAL_TOKENS, metrics.total_tokens
SpanAttributes.TRACELOOP_SPAN_KIND,
TraceloopSpanKindValues.AGENT.value,
)

span.set_status(Status(StatusCode.OK))
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():
input_message = str(args[0])
span.set_attribute(
SpanAttributes.TRACELOOP_ENTITY_INPUT, input_message
)

import time

start_time = time.time()

result = await wrapped(*args, **kwargs)

duration = time.time() - start_time

if hasattr(result, "content") and should_send_prompts():
span.set_attribute(
SpanAttributes.TRACELOOP_ENTITY_OUTPUT, str(result.content)
)

if hasattr(result, "run_id"):
span.set_attribute("agno.run.id", result.run_id)

if hasattr(result, "metrics"):
metrics = result.metrics
if hasattr(metrics, "input_tokens"):
span.set_attribute(
GenAIAttributes.GEN_AI_USAGE_INPUT_TOKENS,
metrics.input_tokens,
)
if hasattr(metrics, "output_tokens"):
span.set_attribute(
GenAIAttributes.GEN_AI_USAGE_OUTPUT_TOKENS,
metrics.output_tokens,
)
if hasattr(metrics, "total_tokens"):
span.set_attribute(
SpanAttributes.LLM_USAGE_TOTAL_TOKENS, metrics.total_tokens
)

span.set_status(Status(StatusCode.OK))

self._duration_histogram.record(
duration,
attributes={
GenAIAttributes.GEN_AI_SYSTEM: "agno",
SpanAttributes.TRACELOOP_SPAN_KIND: TraceloopSpanKindValues.AGENT.value,
},
)

self._duration_histogram.record(
duration,
attributes={
GenAIAttributes.GEN_AI_SYSTEM: "agno",
SpanAttributes.TRACELOOP_SPAN_KIND: TraceloopSpanKindValues.AGENT.value,
},
)
return result

return result
except Exception as e:
span.set_status(Status(StatusCode.ERROR, str(e)))
span.record_exception(e)
raise

except Exception as e:
span.set_status(Status(StatusCode.ERROR, str(e)))
span.record_exception(e)
raise
return async_wrapper()


class _TeamRunWrapper:
Expand Down
Loading