Skip to content

Fix Anthropic usage double-counting cache in input #617

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
elrensmin wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix Anthropic usage double-counting cache in input #617

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
11 changes: 7 additions & 4 deletions src/tau_ai/anthropic.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -734,11 +734,14 @@ def _usage_from_message_start(raw: object) -> Usage:
if isinstance(cache_creation, Mapping)
else None
)
input_tokens = _int_or_none(data.get("input_tokens")) or 0
cache_read = _int_or_none(data.get("cache_read_input_tokens")) or 0
cache_write = _int_or_none(data.get("cache_creation_input_tokens")) or 0
usage = Usage(
input=_int_or_none(data.get("input_tokens")) or 0,
input=max(0, input_tokens - cache_read - cache_write),
output=_int_or_none(data.get("output_tokens")) or 0,
cache_read=_int_or_none(data.get("cache_read_input_tokens")) or 0,
cache_write=_int_or_none(data.get("cache_creation_input_tokens")) or 0,
cache_read=cache_read,
cache_write=cache_write,
cache_write_1h=cache_write_1h,
)
usage.total_tokens = usage.input + usage.output + usage.cache_read + usage.cache_write
Expand All @@ -755,7 +758,7 @@ def _apply_message_delta_usage(usage: Usage | None, raw: object) -> Usage | None
return usage
usage = usage or Usage()
if (value := _int_or_none(raw.get("input_tokens"))) is not None:
usage.input = value
usage.input = max(0, value - usage.cache_read - usage.cache_write)
if (value := _int_or_none(raw.get("output_tokens"))) is not None:
usage.output = value
if (value := _int_or_none(raw.get("cache_read_input_tokens"))) is not None:
Expand Down
4 changes: 2 additions & 2 deletions tests/test_tau_ai.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3130,12 +3130,12 @@ def handler(_request: httpx.Request) -> httpx.Response:
assert isinstance(events[-1], AssistantDoneEvent)
usage = events[-1].message.usage
assert usage is not None
assert usage.input == 100
assert usage.input == 35 # fresh input: 100 - 40 - 25
assert usage.output == 7 # updated by message_delta
assert usage.cache_read == 40
assert usage.cache_write == 25
assert usage.cache_write_1h == 10
assert usage.total_tokens == 172 # 100 + 7 + 40 + 25
assert usage.total_tokens == 107 # 35 + 7 + 40 + 25
assert usage.cost.total == 0


Expand Down