Handle Anthropic usage=None on content-filtered responses

[Copilot]

Anthropic-compatible providers can return stop_reason=content_filter with completion.usage=None. The non-stream query path assumed usage was always present and could crash with AttributeError when extracting token counters.

• Usage extraction hardening

  • Updated ProviderAnthropic._extract_usage to accept Usage | None.
  • Returns zeroed TokenUsage when usage is missing.
  • Normalizes nullable output tokens with or 0 to keep token accounting safe.

• Regression coverage for filtered responses

  • Added a focused test in tests/test_anthropic_kimi_code_provider.py for content-filtered completions with usage=None.
  • Verifies response text is preserved and usage is populated as zero values instead of raising.

def _extract_usage(self, usage: Usage | None) -> TokenUsage:
    if usage is None:
        return TokenUsage()
    return TokenUsage(
        input_other=usage.input_tokens or 0,
        input_cached=usage.cache_read_input_tokens or 0,
        output=usage.output_tokens or 0,
    )

Summary by Sourcery

Handle Anthropic-compatible content-filtered responses that may omit usage data and ensure token accounting remains safe.

Bug Fixes:

• Prevent crashes when Anthropic-compatible providers return content-filtered responses with null usage by defaulting to zeroed token usage.

Tests:

• Add an async test verifying that content-filtered Anthropic responses with missing usage preserve completion text and populate zero-valued usage counters.

[sourcery-ai]

Hey - I've left some high level feedback:

• In the new test, monkeypatch.setattr(anthropic_source, "Message", _FakeMessage) appears unused since provider.client.messages.create is directly patched to return _FakeMessage; consider removing this monkeypatch to keep the test focused and avoid confusion.

Prompt for AI Agents

Please address the comments from this code review:

## Overall Comments
- In the new test, `monkeypatch.setattr(anthropic_source, "Message", _FakeMessage)` appears unused since `provider.client.messages.create` is directly patched to return `_FakeMessage`; consider removing this monkeypatch to keep the test focused and avoid confusion.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}