Spot it. Save it. Free, local-first Chrome extension for images, videos and prompts. Explore 2,000+ curated cases and connect your library to AI Agents via MCP. 提示词导演
JavaScript ai-artai-videochrome-extensionclaude-codecodexhiggsfieldjimenglocal-firstmcpmidjourneymoodboardprompt-engineeringprompt-libraryprompt-managerreference-libraryseedancevisual-bookmarksweb-clipper
import test from "node:test";
import assert from "node:assert/strict";
import { aiConfigurationFromStorage, projectAiRuntime } from "../extension/ai-runtime.js";
import { createComposerSession } from "../extension/composer.js";
import { executeComposerTurnWithService } from "../extension/composer-service.js";
import { readDeepSeekSse } from "../extension/deepseek.js";
import { runComposerToolLoop } from "../extension/composer-tool-loop.js";
function runtime(protocol) {
const endpoint = `https://compatible.example/v1/${protocol === "responses" ? "responses" : "chat/completions"}`;
const configuration = aiConfigurationFromStorage({
aiProviderRegistry: { version: 4, providers: { "custom-text": {
endpoint, protocol, apiKey: "fixture-key", consent: true,
models: { creativePlanning: "account-model" },
discoveredModels: [{ id: "account-model", status: "available", confidence: "declared",
tasks: ["creativePlanning"], inputModalities: ["text", "image"], outputModalities: ["text"] }]
} } },
aiTaskAssignments: { creativePlanning: { providerId: "custom-text", model: "account-model" } }
});
const projected = projectAiRuntime(configuration);
return { endpoint, settings: { ai: projected.aiSettings, vision: projected.visionSettings } };
}
const image = { visualId: "selected-image", dataUrl: "data:image/png;base64,AQID" };
function session() {
return createComposerSession({
aiProfile: { serviceId: "custom-text", model: "account-model" },
referenceSnapshots: [{ entryId: "selected-case", alias: "@参考1", originalText: "原始文字",
imageRefs: [{ visualId: image.visualId, mimeType: "image/png" }] }],
messages: [{ role: "user", type: "request", content: "分析选中的图片" }]
});
}
function streamedResponse(text) {
const bytes = new TextEncoder().encode(text);
// One-byte chunks also split Chinese UTF-8 and CRLF boundaries.
return new Response(new ReadableStream({ start(controller) {
for (const byte of bytes) controller.enqueue(Uint8Array.of(byte));
controller.close();
} }), { headers: { "Content-Type": "text/event-stream" } });
}
for (const protocol of ["chat_completions", "responses"]) {
for (const framing of ["lf", "crlf", "cr", "multiline"]) {
test(`compatible text service preserves selected image and completed ${protocol} reply with ${framing} framing`, async () => {
const { endpoint, settings } = runtime(protocol);
const events = protocol === "responses" ? [
{ type: "response.output_text.delta", delta: "图片分析结果" },
{ type: "response.completed", response: { status: "completed", model: "account-model" } }
] : [
{ choices: [{ delta: { content: "图片分析结果" }, finish_reason: null }] },
{ choices: [{ delta: {}, finish_reason: "stop" }] }
];
const separator = framing === "crlf" ? "\r\n" : framing === "cr" ? "\r" : "\n";
const data = events.map(event => {
const json = JSON.stringify(event, null, framing === "multiline" ? 2 : undefined);
return json.split("\n").map(line => `data: ${line}`).join(separator) + separator + separator;
}).join("");
const calls = [];
const result = await executeComposerTurnWithService({ session: session(), route: "analyze_materials", instruction: "分析参考" }, settings, [image], {
fetchImpl: async (url, options) => { calls.push({ url, body: JSON.parse(options.body) }); return streamedResponse(data); }
});
assert.equal(result.text, "图片分析结果");
assert.equal(calls.length, 1, "one paid request, no protocol retry");
assert.equal(calls[0].url, endpoint);
assert.equal(calls[0].body.model, "account-model");
const content = protocol === "responses" ? calls[0].body.input[0].content : calls[0].body.messages[1].content;
const images = content.filter(part => ["input_image", "image_url"].includes(part.type));
assert.equal(images.length, 1);
assert.equal(protocol === "responses" ? images[0].image_url : images[0].image_url.url, image.dataUrl);
});
}
test(`compatible ${protocol} does not save a reply whose terminal event was cut off`, async () => {
const { settings } = runtime(protocol);
const delta = protocol === "responses" ? { type: "response.output_text.delta", delta: "部分结果" }
: { choices: [{ delta: { content: "部分结果" } }] };
const terminal = protocol === "responses" ? { type: "response.completed", response: { status: "completed" } }
: { choices: [{ delta: {}, finish_reason: "stop" }] };
let calls = 0;
await assert.rejects(executeComposerTurnWithService({ session: session(), route: "analyze_materials" }, settings, [image], {
fetchImpl: async () => { calls++; return streamedResponse(`data: ${JSON.stringify(delta)}\n\ndata: ${JSON.stringify(terminal).slice(0, -1)}`); }
}), /完整|中断/);
assert.equal(calls, 1);
});
}
test("text generation reads standard CR-only events and the required completion marker", async () => {
const events = [
{ choices: [{ delta: { content: "中文回复" } }] },
{ choices: [{ delta: {}, finish_reason: "stop" }] }
];
const response = streamedResponse(events.map(event => `data: ${JSON.stringify(event)}\r\r`).join("") + "data: [DONE]\r\r");
assert.equal((await readDeepSeekSse(response)).content, "中文回复");
});
for (const complete of [true, false]) {
test(`tool execution requires a complete CR-only event: ${complete}`, async () => {
let requests = 0;
let executions = 0;
const event = { choices: [{ delta: { tool_calls: [{ index: 0, id: "call-1", type: "function",
function: { name: "search_cases", arguments: '{"query":"雨夜"}' } }] }, finish_reason: "tool_calls" }] };
const run = runComposerToolLoop({
body: { model: "fixture", messages: [{ role: "user", content: "查资料" }] }, protocol: "chat_completions", maxCharacters: 10000,
request: async () => {
requests++;
return requests === 1 ? streamedResponse(`data: ${complete ? JSON.stringify(event) + "\r\r" : JSON.stringify(event).slice(0, -1)}`)
: new Response(JSON.stringify({ choices: [{ message: { content: "找到资料" }, finish_reason: "stop" }] }), { headers: { "Content-Type": "application/json" } });
},
runtime: { specs: [{ name: "search_cases", parameters: { type: "object", properties: { query: { type: "string" } } } }],
execute: async (_name, args) => { executions++; assert.equal(args.query, "雨夜"); return { data: { count: 1 } }; } }
});
if (complete) {
assert.equal((await run).content, "找到资料");
assert.equal(executions, 1);
assert.equal(requests, 2);
} else {
await assert.rejects(run, /中断|JSON/);
assert.equal(executions, 0);
assert.equal(requests, 1);
}
});
}