).rejects.toThrow(/did not deliver a first event within 1ms after HTTP streaming headers/);

});

it("clamps Responses cached prompt usage at zero", async () => {

const model = createAzureResponsesModel();

const output = createResponsesAssistantOutput(model);

await __testing.processResponsesStream(

streamChunks([

{

type: "response.completed",

response: {

id: "resp-cache-overflow",

status: "completed",

usage: {

input_tokens: 2,

output_tokens: 5,

total_tokens: 7,

input_tokens_details: { cached_tokens: 4 },

},

},

},

]),

output,

{ push: vi.fn() },

model,

);

expectRecordFields(output.usage, {

input: 0,

output: 5,

cacheRead: 4,

totalTokens: 9,

});

});

it("summarizes model payload tools with full names when requested", () => {

const previous = process.env.OPENCLAW_DEBUG_MODEL_PAYLOAD;

process.env.OPENCLAW_DEBUG_MODEL_PAYLOAD = "tools";

| undefined;

if (usage) {

const cachedTokens = usage.input_tokens_details?.cached_tokens || 0;

const inputTokens = usage.input_tokens || 0;

const outputTokens = usage.output_tokens || 0;

const input = Math.max(0, inputTokens - cachedTokens);

output.usage = {

input: (usage.input_tokens || 0) - cachedTokens,

output: usage.output_tokens || 0,

input,

output: outputTokens,

cacheRead: cachedTokens,

cacheWrite: 0,

totalTokens: usage.total_tokens || 0,

totalTokens: input + outputTokens + cachedTokens,

cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },

};

}