Skip to content

feat(llms/anthropic): Add support for RawMessageDeltaEvent in streaming #20206

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
merged 6 commits into from
Nov 10, 2025
+350 −1
Merged

feat(llms/anthropic): Add support for RawMessageDeltaEvent in streaming #20206

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
88555d1
feat(llms/anthropic): Add support for RawMessageDeltaEvent in streaming
zhangzhefang-github Nov 2, 2025
7ac1df0
fix: Ensure stop_reason is captured by yielding on RawMessageDeltaEvent
zhangzhefang-github Nov 4, 2025
7398cd1
Merge branch 'main' into feat/anthropic-raw-message-delta-event
logan-markewich Nov 7, 2025
b828018
vbump
logan-markewich Nov 7, 2025
678e019
test: Add mock tests for streaming metadata capture (coverage >50%)
zhangzhefang-github Nov 8, 2025
734b7ad
fix lint
logan-markewich Nov 9, 2025
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
93 changes: 93 additions & 0 deletions llama-index-integrations/llms/llama-index-llms-anthropic/llama_index/llms/anthropic/base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,6 +59,9 @@
RawContentBlockDeltaEvent,
RawContentBlockStartEvent,
RawContentBlockStopEvent,
RawMessageDeltaEvent,
RawMessageStartEvent,
RawMessageStopEvent,
TextBlock,
TextDelta,
ThinkingBlock,
Expand Down Expand Up @@ -462,6 +465,9 @@ def gen() -> Generator[AnthropicChatResponse, None, None]:
cur_citations: List[Dict[str, Any]] = []
tracked_citations: Set[str] = set()
role = MessageRole.ASSISTANT
# Track usage metadata and stop_reason from RawMessage events
usage_metadata: Dict[str, Any] = {}
stop_reason: Optional[str] = None
for r in response:
if isinstance(r, (ContentBlockDeltaEvent, RawContentBlockDeltaEvent)):
if isinstance(r.delta, TextDelta):
Expand Down Expand Up @@ -550,6 +556,10 @@ def gen() -> Generator[AnthropicChatResponse, None, None]:
message=ChatMessage(
role=role,
blocks=content,
additional_kwargs={
"usage": usage_metadata if usage_metadata else None,
"stop_reason": stop_reason,
},
),
citations=cur_citations,
delta=content_delta,
Expand Down Expand Up @@ -584,11 +594,49 @@ def gen() -> Generator[AnthropicChatResponse, None, None]:
message=ChatMessage(
role=role,
blocks=content,
additional_kwargs={
"usage": usage_metadata if usage_metadata else None,
"stop_reason": stop_reason,
},
),
citations=cur_citations,
delta="",
raw=dict(r),
)
elif isinstance(r, RawMessageStartEvent):
# Capture initial usage metadata from message_start
if hasattr(r.message, "usage") and r.message.usage:
usage_metadata = {
"input_tokens": r.message.usage.input_tokens,
"output_tokens": r.message.usage.output_tokens,
}
elif isinstance(r, RawMessageDeltaEvent):
# Update usage metadata and capture stop_reason from message_delta
if hasattr(r, "usage") and r.usage:
usage_metadata = {
"input_tokens": r.usage.input_tokens,
"output_tokens": r.usage.output_tokens,
}
if hasattr(r, "delta") and hasattr(r.delta, "stop_reason"):
stop_reason = r.delta.stop_reason

# Yield a final chunk with updated metadata including stop_reason
yield AnthropicChatResponse(
message=ChatMessage(
role=role,
blocks=content,
additional_kwargs={
"usage": usage_metadata if usage_metadata else None,
"stop_reason": stop_reason,
},
),
citations=cur_citations,
delta="",
raw=dict(r),
)
elif isinstance(r, RawMessageStopEvent):
# Final event - no additional data to capture
pass

return gen()

Expand Down Expand Up @@ -664,6 +712,9 @@ async def gen() -> ChatResponseAsyncGen:
cur_citations: List[Dict[str, Any]] = []
tracked_citations: Set[str] = set()
role = MessageRole.ASSISTANT
# Track usage metadata and stop_reason from RawMessage events
usage_metadata: Dict[str, Any] = {}
stop_reason: Optional[str] = None
async for r in response:
if isinstance(r, (ContentBlockDeltaEvent, RawContentBlockDeltaEvent)):
if isinstance(r.delta, TextDelta):
Expand Down Expand Up @@ -752,6 +803,10 @@ async def gen() -> ChatResponseAsyncGen:
message=ChatMessage(
role=role,
blocks=content,
additional_kwargs={
"usage": usage_metadata if usage_metadata else None,
"stop_reason": stop_reason,
},
),
citations=cur_citations,
delta=content_delta,
Expand Down Expand Up @@ -786,11 +841,49 @@ async def gen() -> ChatResponseAsyncGen:
message=ChatMessage(
role=role,
blocks=content,
additional_kwargs={
"usage": usage_metadata if usage_metadata else None,
"stop_reason": stop_reason,
},
),
citations=cur_citations,
delta="",
raw=dict(r),
)
elif isinstance(r, RawMessageStartEvent):
# Capture initial usage metadata from message_start
if hasattr(r.message, "usage") and r.message.usage:
usage_metadata = {
"input_tokens": r.message.usage.input_tokens,
"output_tokens": r.message.usage.output_tokens,
}
elif isinstance(r, RawMessageDeltaEvent):
# Update usage metadata and capture stop_reason from message_delta
if hasattr(r, "usage") and r.usage:
usage_metadata = {
"input_tokens": r.usage.input_tokens,
"output_tokens": r.usage.output_tokens,
}
if hasattr(r, "delta") and hasattr(r.delta, "stop_reason"):
stop_reason = r.delta.stop_reason

# Yield a final chunk with updated metadata including stop_reason
yield AnthropicChatResponse(
message=ChatMessage(
role=role,
blocks=content,
additional_kwargs={
"usage": usage_metadata if usage_metadata else None,
"stop_reason": stop_reason,
},
),
citations=cur_citations,
delta="",
raw=dict(r),
)
elif isinstance(r, RawMessageStopEvent):
# Final event - no additional data to capture
pass

return gen()

Expand Down
2 changes: 1 addition & 1 deletion llama-index-integrations/llms/llama-index-llms-anthropic/pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@ dev = [

[project]
name = "llama-index-llms-anthropic"
version = "0.10.1"
version = "0.10.2"
description = "llama-index llms anthropic integration"
authors = [{name = "Your Name", email = "you@example.com"}]
requires-python = ">=3.9,<4.0"
Expand Down
Loading