Skip to content

feat(litellm): Support span streaming #6317

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

Open
alexander-alderman-webb wants to merge 3 commits into
base: master
Choose a base branch
from
+253 −43
Open

feat(litellm): Support span streaming #6317

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
49ec553
feat(litellm): Support span streaming
alexander-alderman-webb May 21, 2026
5d8de73
Merge branch 'master' into webb/litellm/span-first
alexander-alderman-webb May 28, 2026
1690b64
test
alexander-alderman-webb May 28, 2026
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
52 changes: 35 additions & 17 deletions sentry_sdk/integrations/litellm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,10 @@
from sentry_sdk.consts import SPANDATA
from sentry_sdk.integrations import DidNotEnable, Integration
from sentry_sdk.scope import should_send_default_pii
from sentry_sdk.tracing_utils import (
has_span_streaming_enabled,
should_truncate_gen_ai_input,
)
from sentry_sdk.utils import event_from_exception

if TYPE_CHECKING:
Expand Down Expand Up @@ -68,7 +72,8 @@ def _convert_message_parts(messages: "List[Dict[str, Any]]") -> "List[Dict[str,

def _input_callback(kwargs: "Dict[str, Any]") -> None:
"""Handle the start of a request."""
integration = sentry_sdk.get_client().get_integration(LiteLLMIntegration)
client = sentry_sdk.get_client()
integration = client.get_integration(LiteLLMIntegration)

if integration is None:
return
Expand All @@ -88,16 +93,29 @@ def _input_callback(kwargs: "Dict[str, Any]") -> None:
operation = "chat"

# Start a new span/transaction
span = get_start_span_function()(
op=(
consts.OP.GEN_AI_CHAT
if operation == "chat"
else consts.OP.GEN_AI_EMBEDDINGS
),
name=f"{operation} {model}",
origin=LiteLLMIntegration.origin,
)
span.__enter__()
if has_span_streaming_enabled(client.options):
span = sentry_sdk.traces.start_span(
name=f"{operation} {model}",
attributes={
"sentry.op": (
consts.OP.GEN_AI_CHAT
if operation == "chat"
else consts.OP.GEN_AI_EMBEDDINGS
Copy link
Copy Markdown
Member

@ericapisani ericapisani May 28, 2026

Choose a reason for hiding this comment

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

I know this existed in the previous code, but this fallback seems like an interesting choice. I would've expected the inverse where chat is the fallback since I suspect it's the most common operation that would be performed with an AI provider 🤔

),
"sentry.origin": LiteLLMIntegration.origin,
},
)
else:
span = get_start_span_function()(
op=(
consts.OP.GEN_AI_CHAT
if operation == "chat"
else consts.OP.GEN_AI_EMBEDDINGS
),
name=f"{operation} {model}",
origin=LiteLLMIntegration.origin,
)
span.__enter__()

# Store span for later
_get_metadata_dict(kwargs)["_sentry_span"] = span
Expand All @@ -121,9 +139,9 @@ def _input_callback(kwargs: "Dict[str, Any]") -> None:
)
client = sentry_sdk.get_client()
messages_data = (
input_list
if client.options.get("stream_gen_ai_spans", False)
else truncate_and_annotate_embedding_inputs(input_list, span, scope)
truncate_and_annotate_embedding_inputs(input_list, span, scope)
if should_truncate_gen_ai_input(client.options)
else input_list
)
if messages_data is not None:
set_data_normalized(
Expand All @@ -140,9 +158,9 @@ def _input_callback(kwargs: "Dict[str, Any]") -> None:
scope = sentry_sdk.get_current_scope()
messages = _convert_message_parts(messages)
messages_data = (
messages
if client.options.get("stream_gen_ai_spans", False)
else truncate_and_annotate_messages(messages, span, scope)
truncate_and_annotate_messages(messages, span, scope)
if should_truncate_gen_ai_input(client.options)
else messages
)
if messages_data is not None:
set_data_normalized(
Expand Down
Loading
Loading