import { test } from "bun:test"; import assert from "node:assert/strict"; import { buildOpenAiToolLoopInput, buildOpenAiReasoningParam, buildOpenAiTemperatureParam, formatProviderResponseShape, summarizeProviderRawContent } from "./serviceShared.ts";

test("buildOpenAiTemperatureParam omits temperature for gpt-5 dot-version models", () => { assert.deepEqual(buildOpenAiTemperatureParam("gpt-5.4-mini", 0.7), {}); assert.deepEqual(buildOpenAiTemperatureParam("gpt-5.4", 0.2), {}); assert.deepEqual(buildOpenAiTemperatureParam("gpt-5-mini", 0.2), {}); });

test("buildOpenAiTemperatureParam keeps temperature for non-gpt-5 families", () => { assert.deepEqual(buildOpenAiTemperatureParam("gpt-4o-mini", 0.6), { temperature: 0.6 }); });

test("buildOpenAiReasoningParam applies to gpt-5 dot-version models", () => { assert.deepEqual(buildOpenAiReasoningParam("gpt-5.4-mini", "minimal"), { reasoning: { effort: "low" } }); assert.deepEqual(buildOpenAiReasoningParam("gpt-4o-mini", "minimal"), {}); });

test("buildOpenAiReasoningParam accepts none and xhigh efforts", () => { assert.deepEqual(buildOpenAiReasoningParam("gpt-5.4-mini", "none"), { reasoning: { effort: "none" } }); assert.deepEqual(buildOpenAiReasoningParam("gpt-5.4-mini", "xhigh"), { reasoning: { effort: "xhigh" } }); });

test("buildOpenAiToolLoopInput omits synthetic assistant ids for OpenAI responses compatibility", () => { const input = buildOpenAiToolLoopInput([ { role: "user", content: "play some minecraft music" }, { role: "assistant", content: [ { type: "text", text: "Sure, I'll start that." }, { type: "tool_call", id: "call_music_play", name: "music_play", input: { query: "minecraft soundtrack" } } ] }, { role: "user", content: [ { type: "tool_result", toolCallId: "call_music_play", content: "Playing C418 - Sweden" } ] } ]);

assert.equal(input[1]?.type, "message"); assert.equal(Object.prototype.hasOwnProperty.call(input[1] || {}, "id"), false);

assert.equal(input[2]?.type, "function_call"); assert.equal(Object.prototype.hasOwnProperty.call(input[2] || {}, "id"), false); });

test("summarizeProviderRawContent captures OpenAI response output shape without raw text", () => { const summary = summarizeProviderRawContent([ { id: "msg_1", type: "message", role: "assistant", status: "completed", content: [ { type: "output_text", text: "done" } ] }, { id: "fc_1", type: "function_call", name: "spawn_code_worker", arguments: "{"task":"create a todo app"}", status: "completed" } ]);

assert.equal(summary.shape, "array"); assert.equal(summary.length, 2); assert.deepEqual(summary.itemTypes, { message: 1, function_call: 1 }); assert.deepEqual(summary.contentPartTypes, { output_text: 1 }); assert.equal(summary.functionCallCount, 1); assert.deepEqual(summary.functionCallNames, ["spawn_code_worker"]); assert.equal(summary.functionCallArgumentPresent, true); assert.equal(summary.textChars, 4); });

test("formatProviderResponseShape includes concise stream extraction diagnostics", () => { const summary = summarizeProviderRawContent([]); const shape = formatProviderResponseShape(summary, { streamDeltaTextChars: 0, streamDoneTextChars: 0, extractedTextChars: 0, finalOutputItemCount: 0 });

assert.equal(shape, "raw=array[0] delta=0 done=0 extracted=0 finalOut=0"); });

test("formatProviderResponseShape includes recovered streamed tool calls", () => { const summary = summarizeProviderRawContent([ { id: "fc_1", type: "function_call", name: "spawn_code_worker", arguments: "{"task":"create a todo app"}" } ]); const shape = formatProviderResponseShape(summary, { streamDeltaTextChars: 0, streamDoneTextChars: 0, extractedTextChars: 0, finalOutputItemCount: 0, streamRecoveredToolCallCount: 1 });

assert.equal(shape.includes("streamTools=1"), true); });