feat: add token usage tracking to OpenAI adapter

[sestinj]

Summary

• Added token usage tracking to the OpenAI adapter to match the existing implementations in Anthropic and Gemini adapters
• Modified the streaming response handler to properly collect and emit usage information
• Verified all three providers (OpenAI, Anthropic, Gemini) now consistently track and report token usage

Changes

• OpenAI Adapter: Updated chatCompletionStream method to handle usage chunks that arrive at the end of the stream
• Tests: Added comprehensive test coverage for token usage tracking across all three providers
• Verification: All existing tests pass with the expectUsage: true flag

Test Plan

• Run existing test suite with API keys
• Verify OpenAI adapter properly tracks usage in streaming responses
• Verify Anthropic adapter continues to track usage correctly
• Verify Gemini adapter continues to track usage correctly
• Add unit tests for token usage tracking

Linear Issue

CON-3935

🤖 Generated with Claude Code

---

Summary by cubic

Adds token usage tracking to the OpenAI adapter and defers the usage event until after all streamed content. Aligns OpenAI with Anthropic and Gemini so all providers report prompt, completion, and total tokens (CON-3935).

• New Features
  • Update OpenAI chatCompletionStream to buffer the usage chunk and emit it last; non-streaming passes usage through as-is.
  • Add tests for usage tracking across OpenAI, Anthropic, and Gemini (streaming and non-streaming).

[cubic-dev-ai]

1 issue found across 3 files

Prompt for AI agents (all 1 issues)

Understand the root cause of the following 1 issues and fix them.


<file name="packages/openai-adapters/src/test/token-usage.test.ts">

<violation number="1" location="packages/openai-adapters/src/test/token-usage.test.ts:115">
Overwriting `global.fetch` without restoring leaves the mock active for subsequent tests. Please store the original fetch and restore it in afterEach/afterAll (or use `vi.spyOn`) so other suites keep the real implementation.</violation>
</file>

_{React with 👍 or 👎 to teach cubic. Mention @cubic-dev-ai to give feedback, ask questions, or re-run the review.}

[cubic-dev-ai]

Overwriting global.fetch without restoring leaves the mock active for subsequent tests. Please store the original fetch and restore it in afterEach/afterAll (or use vi.spyOn) so other suites keep the real implementation.

Prompt for AI agents

Address the following comment on packages/openai-adapters/src/test/token-usage.test.ts at line 115:

<comment>Overwriting `global.fetch` without restoring leaves the mock active for subsequent tests. Please store the original fetch and restore it in afterEach/afterAll (or use `vi.spyOn`) so other suites keep the real implementation.</comment>

<file context>
@@ -0,0 +1,353 @@
+      }),
+    };
+
+    global.fetch = vi.fn().mockResolvedValue(mockResponse);
+
+    const api = new AnthropicApi({ apiKey: &quot;test&quot;, provider: &quot;anthropic&quot; });
</file context>