Skip to content

The agent that grows with you

test(e2e): fake Responses server sends cache_write_tokens the openai 2.54 SDK requires

Tekniume3cc65d767e3
diff --git a/tests/fakes/providers/openai_responses.py b/tests/fakes/providers/openai_responses.py
index 8472187c6b2..191ce514866 100644
--- a/tests/fakes/providers/openai_responses.py
+++ b/tests/fakes/providers/openai_responses.py
@@ -93,6 +93,7 @@ class Turn:
input_tokens: int = 100
output_tokens: int = 20
cached_tokens: int = 0
+ cache_write_tokens: int = 0
drop_after_events: int | None = None # close the socket after N events (no terminal event)
@@ -260,7 +261,8 @@ def turn_events(turn: Turn) -> list[dict[str, Any]]:
usage = ResponseUsage(
input_tokens=turn.input_tokens, output_tokens=turn.output_tokens,
total_tokens=turn.input_tokens + turn.output_tokens,
- input_tokens_details=InputTokensDetails(cached_tokens=turn.cached_tokens),
+ input_tokens_details=InputTokensDetails(cached_tokens=turn.cached_tokens,
+ cache_write_tokens=turn.cache_write_tokens),
output_tokens_details=OutputTokensDetails(reasoning_tokens=0))
events.append(ResponseCompletedEvent(response=_response(resp_id, "completed", built, usage),
sequence_number=seq(), type="response.completed"))