Skip to content

Never stop coding. Free MIT AI gateway: one endpoint, 375 providers (150+ free), 1,400+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 900+ contributors

fix: extract flat cached_tokens/reasoning_tokens from OpenAI-compatible usage objects (#2350)

Tyler Fordd6d585e20099
diff --git a/open-sse/handlers/usageExtractor.ts b/open-sse/handlers/usageExtractor.ts
index e2e87983a9..92affd1c39 100644
--- a/open-sse/handlers/usageExtractor.ts
+++ b/open-sse/handlers/usageExtractor.ts
@@ -25,10 +25,12 @@ export function extractUsageFromResponse(responseBody, provider) {
cached_tokens:
responseBody.usage.prompt_tokens_details?.cached_tokens ??
responseBody.usage.input_tokens_details?.cached_tokens ??
- responseBody.usage.prompt_cache_hit_tokens,
+ responseBody.usage.prompt_cache_hit_tokens ??
+ responseBody.usage.cached_tokens,
reasoning_tokens:
responseBody.usage.completion_tokens_details?.reasoning_tokens ??
- responseBody.usage.output_tokens_details?.reasoning_tokens,
+ responseBody.usage.output_tokens_details?.reasoning_tokens ??
+ responseBody.usage.reasoning_tokens,
};
}
diff --git a/open-sse/utils/usageTracking.ts b/open-sse/utils/usageTracking.ts
index 15e7d681d4..38913a2263 100644
--- a/open-sse/utils/usageTracking.ts
+++ b/open-sse/utils/usageTracking.ts
@@ -366,10 +366,12 @@ export function extractUsage(chunk) {
cached_tokens:
chunk.usage.prompt_tokens_details?.cached_tokens ??
chunk.usage.input_tokens_details?.cached_tokens ??
- chunk.usage.prompt_cache_hit_tokens,
+ chunk.usage.prompt_cache_hit_tokens ??
+ chunk.usage.cached_tokens,
reasoning_tokens:
chunk.usage.completion_tokens_details?.reasoning_tokens ??
- chunk.usage.output_tokens_details?.reasoning_tokens,
+ chunk.usage.output_tokens_details?.reasoning_tokens ??
+ chunk.usage.reasoning_tokens,
});
}
diff --git a/tests/unit/usage-extractor.test.ts b/tests/unit/usage-extractor.test.ts
index 4d64bf6eca..6c7bdf5273 100644
--- a/tests/unit/usage-extractor.test.ts
+++ b/tests/unit/usage-extractor.test.ts
@@ -292,3 +292,43 @@ test("extractUsage reads OpenAI streaming chunk with prompt_tokens_details", ()
assert.equal(usage.cached_tokens, 50);
assert.equal(usage.reasoning_tokens, 20);
});
+
+// ── Flat field extraction tests (Xiaomi MiMo-style providers) ──
+
+test("extractUsageFromResponse reads flat cached_tokens and reasoning_tokens from OpenAI-compatible usage", () => {
+ const usage = extractUsageFromResponse(
+ {
+ usage: {
+ prompt_tokens: 258,
+ completion_tokens: 50,
+ total_tokens: 308,
+ cached_tokens: 192,
+ reasoning_tokens: 49,
+ },
+ },
+ "xiaomi-mimo"
+ );
+
+ assert.deepEqual(usage, {
+ prompt_tokens: 258,
+ completion_tokens: 50,
+ cached_tokens: 192,
+ reasoning_tokens: 49,
+ });
+});
+
+test("extractUsage reads flat cached_tokens and reasoning_tokens from streaming chunk", () => {
+ const usage = extractUsage({
+ choices: [{ delta: {}, finish_reason: "stop" }],
+ usage: {
+ prompt_tokens: 258,
+ completion_tokens: 50,
+ total_tokens: 308,
+ cached_tokens: 192,
+ reasoning_tokens: 49,
+ },
+ });
+
+ assert.equal(usage.cached_tokens, 192);
+ assert.equal(usage.reasoning_tokens, 49);
+});