chore: checkpoint pi lab rust integration work

This commit is contained in:
Agent Board
2026-07-11 20:39:18 +08:00
parent d47f6447fd
commit d16eccfe10
39 changed files with 4843 additions and 455 deletions
+56 -7
View File
@@ -16,7 +16,7 @@ const WORKSPACE_ID = process.env.MNOTE_E2E_WORKSPACE_ID || "local-ws:mnote-e2e:m
const ROOT_PATH = process.env.MNOTE_E2E_ROOT_PATH || "/mnt/Data1T/Mnote_data/users/mnote-e2e/workspaces/my-space";
const ROOT_URI = process.env.MNOTE_E2E_ROOT_URI || `file://${ROOT_PATH}`;
const MODEL_PROVIDER = process.env.MNOTE_PI_REAL_MODEL_PROVIDER || "omniroute";
const MODEL_ID = process.env.MNOTE_PI_REAL_MODEL_ID || "freefirst";
const MODEL_ID = process.env.MNOTE_PI_REAL_MODEL_ID || "gpt-5.4-mini";
const MARKER = `PI_REAL_LIGHTRAG_CARBOXY_OK_${STAMP}`;
const CHROMIUM_EXECUTABLE = process.env.PLAYWRIGHT_CHROMIUM_EXECUTABLE
|| (fs.existsSync("/usr/bin/chromium-browser") ? "/usr/bin/chromium-browser" : "")
@@ -185,15 +185,60 @@ async function expandToolTimelines(page) {
}
function readSessionJsonl(sessionDir) {
const files = fs.readdirSync(sessionDir)
.filter((name) => name.endsWith(".jsonl"))
.map((name) => path.join(sessionDir, name))
.sort();
const files = [];
const walk = (dir) => {
for (const entry of fs.readdirSync(dir, { withFileTypes: true })) {
const full = path.join(dir, entry.name);
if (entry.isDirectory()) walk(full);
else if (entry.isFile() && entry.name.endsWith(".jsonl")) files.push(full);
}
};
walk(sessionDir);
files.sort();
const sessionFile = files[files.length - 1];
assert(sessionFile, `session jsonl not found in ${sessionDir}`);
return { sessionFile, raw: fs.readFileSync(sessionFile, "utf8") };
}
function sessionHasKnowledgeRagEvidence(raw) {
return String(raw || "")
.split("\n")
.filter(Boolean)
.some((line) => {
try {
const entry = JSON.parse(line);
const message = entry && entry.message;
if (
!message
|| message.role !== "toolResult"
|| !/mnote_knowledge_rag_query|mnote\.knowledge_rag\.query/.test(String(message.toolName || ""))
) {
return false;
}
const result = message.details && message.details.result;
return Boolean(
result
&& ((Array.isArray(result.references) && result.references.length)
|| (Array.isArray(result.citations) && result.citations.length)
|| (Array.isArray(result.uiCitations) && result.uiCitations.length)),
);
} catch {
return false;
}
});
}
async function waitForSessionKnowledgeRagEvidence(sessionDir, timeoutMs = 10000) {
const started = Date.now();
let snapshot = readSessionJsonl(sessionDir);
while (Date.now() - started < timeoutMs) {
if (sessionHasKnowledgeRagEvidence(snapshot.raw)) return snapshot;
await new Promise((resolve) => setTimeout(resolve, 200));
snapshot = readSessionJsonl(sessionDir);
}
return snapshot;
}
function extractAssistantText(text) {
return text
.replace(/\s+/g, " ")
@@ -273,6 +318,8 @@ async function main() {
.last();
await markerLocator.waitFor({ state: "visible", timeout: TIMEOUT });
await expandToolTimelines(page);
const citationLocator = page.locator("[data-page-ai-pi-lab-citation]").first();
await citationLocator.waitFor({ state: "visible", timeout: 10000 }).catch(() => null);
const toolTimeline = page.locator("[data-page-ai-pi-lab-tool-timeline], [data-page-ai-pi-tool-timeline]").first();
await toolTimeline.scrollIntoViewIfNeeded({ timeout: TIMEOUT }).catch(() => null);
await page.screenshot({ path: path.join(OUT, "02-real-pi-lightrag-tool-card.png"), fullPage: false });
@@ -284,17 +331,19 @@ async function main() {
const assistantText = extractAssistantText((await markerLocator.textContent({ timeout: TIMEOUT })) || "");
result.answerText = assistantText;
const toolText = (await page.locator("[data-page-ai-pi-lab-tool-timeline], [data-page-ai-pi-tool-timeline]").allTextContents()).join("\n");
const sessionJsonl = readSessionJsonl(session.piSessionDir);
const sessionJsonl = await waitForSessionKnowledgeRagEvidence(session.piSessionDir);
result.session.sessionFile = sessionJsonl.sessionFile;
result.checks.toolCardVisible = /LightRAG|knowledge_rag|mnote_knowledge_rag_query|mnote\.knowledge_rag\.query/i.test(toolText);
result.checks.queryToolCalled = /mnote_knowledge_rag_query|mnote\.knowledge_rag\.query/.test(sessionJsonl.raw);
result.checks.referencesReturned = /references|citations|citationMarkdown|displayQuote/.test(sessionJsonl.raw);
result.checks.referencesReturned = sessionHasKnowledgeRagEvidence(sessionJsonl.raw);
result.checks.citationChipsVisible = await page.locator("[data-page-ai-pi-lab-citation]").count() > 0;
result.checks.answerHasFiveItems = (assistantText.match(/(^|\s)([1-5]([.、.]|️⃣)|[-*]\s)/g) || []).length >= 5 || /5\s*种|五种|5\s*種|五種/.test(assistantText);
result.checks.answerMentionsEvidence = /(\[[0-9]{3,4}\]|苯甲酰溴甲酯|碳酸二|硫酸二甲酯|磷酸三甲酯|引用|原文)/.test(assistantText);
result.checks.noReasoningLeakInToolTimeline = !/reasoning_content|\"thinking\"/i.test(toolText);
assert(result.checks.toolCardVisible, "UI 未显示 LightRAG/MNote RAG 工具卡");
assert(result.checks.queryToolCalled, "Pi session JSONL 未记录 LightRAG 查询工具调用");
assert(result.checks.referencesReturned, "Pi session JSONL 未包含 LightRAG references/citations");
assert(result.checks.citationChipsVisible, "Pi 最终回答未显示可点击的 LightRAG 引用");
assert(result.checks.answerHasFiveItems, "Pi 最终回答未明显列出 5 项");
assert(result.checks.answerMentionsEvidence, "Pi 最终回答未明显包含引用依据");
assert(result.checks.noReasoningLeakInToolTimeline, "工具时间线不应泄漏 thinking/reasoning 字段");