Skip to content

Fix BedrockConverse token count in usage to match OpenAI's format #18383

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
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fix BedrockConverse token count in usage to match OpenAI's format #18383

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ddcf0d2
feat: update token count in usage to match OpenAI's format
ArmandBriere Apr 6, 2025
1932f79
fix: apply new token count fix to every chat function for stream and …
ArmandBriere Apr 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
20 changes: 16 additions & 4 deletions ...grations/llms/llama-index-llms-bedrock-converse/llama_index/llms/bedrock_converse/base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -325,6 +325,10 @@ def chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
response
)

dict_response = dict(response)
# Add Bedrock's token count to usage dict to match OpenAI's format
dict_response["usage"] = self._get_response_token_counts(dict_response)

return ChatResponse(
message=ChatMessage(
role=MessageRole.ASSISTANT,
Expand All @@ -335,7 +339,7 @@ def chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
"status": status,
},
),
raw=dict(response),
raw=dict_response,
additional_kwargs=self._get_response_token_counts(dict(response)),
)

Expand Down Expand Up @@ -367,6 +371,10 @@ def stream_chat(
**all_kwargs,
)

dict_response = dict(response)
# Add Bedrock's token count to usage dict to match OpenAI's format
dict_response["usage"] = self._get_response_token_counts(dict_response)
Copy link
Collaborator

Choose a reason for hiding this comment

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

You'll actually need to get the usage on each chunk in the stream? I think? You might need to debug whats beining emitted in the chunk stream below to get the usage properly

Copy link
Collaborator

Choose a reason for hiding this comment

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

It can't know the complete usage until the stream is finished, for example


def gen() -> ChatResponseGen:
content = {}
role = MessageRole.ASSISTANT
Expand All @@ -392,7 +400,7 @@ def gen() -> ChatResponseGen:
},
),
delta=content_delta.get("text", ""),
raw=response,
raw=dict_response,
additional_kwargs=self._get_response_token_counts(
dict(response)
),
Expand All @@ -417,7 +425,7 @@ def gen() -> ChatResponseGen:
"status": status,
},
),
raw=response,
raw=dict_response,
additional_kwargs=self._get_response_token_counts(
dict(response)
),
Expand Down Expand Up @@ -458,6 +466,10 @@ async def achat(
response
)

dict_response = dict(response)
# Add Bedrock's token count to usage dict to match OpenAI's format
dict_response["usage"] = self._get_response_token_counts(dict_response)

return ChatResponse(
message=ChatMessage(
role=MessageRole.ASSISTANT,
Expand All @@ -468,7 +480,7 @@ async def achat(
"status": status,
},
),
raw=dict(response),
raw=dict_response,
additional_kwargs=self._get_response_token_counts(dict(response)),
)

Expand Down