fix(web-probe): bind new session command to mutation response (#681)

Co-authored-by: Codex <codex@noreply.local>

scripts/src/hwlab-node-web-observe-runner-source.ts

@@ -684,21 +684,47 @@ async function createSessionFromUi() {
   const before = await workbenchSessionSnapshot();
   const create = page.locator("#session-create").first();
   await create.waitFor({ state: "visible", timeout: 15000 });
+  const createResponsePromise = page.waitForResponse((response) => {
+    const request = response.request();
+    if (request.method().toUpperCase() !== "POST") return false;
+    try {
+      return new URL(response.url()).pathname === "/v1/agent/sessions";
+    } catch {
+      return false;
+    }
+  }, { timeout: 45000 }).catch((error) => ({ waitError: errorSummary(error) }));
   await create.click();
-  await page.waitForFunction((initial) => {
+  const createResponse = await createResponsePromise;
+  if (createResponse?.waitError) throw new Error("newSession did not observe POST /v1/agent/sessions response after click: " + (createResponse.waitError.message || createResponse.waitError.name || "timeout"));
+  const createStatus = createResponse.status();
+  let createPayload = null;
+  let createPayloadError = null;
+  try {
+    createPayload = await createResponse.json();
+  } catch (error) {
+    createPayloadError = errorSummary(error);
+  }
+  const createdSessionId = sessionIdFromAgentSessionPayload(createPayload);
+  if (createStatus < 200 || createStatus >= 300 || !createdSessionId) {
+    const error = new Error("newSession did not receive an authoritative session id from POST /v1/agent/sessions");
+    error.details = { status: createStatus, statusText: createResponse.statusText(), responseParsed: createPayload !== null, responseParseError: createPayloadError, createdSessionId, valuesRedacted: true };
+    throw error;
+  }
+  await page.waitForFunction((expectedSessionId) => {
     const activeTab = document.querySelector(".session-tab[data-active='true'], .session-tab[aria-selected='true']");
     const sessionId = activeTab?.getAttribute("data-session-id") || "";
+    const routeMatch = window.location.pathname.match(/\/workbench\/sessions\/([^/]+)/u) || window.location.pathname.match(/\/workspace\/sessions\/([^/]+)/u);
+    const routeSessionId = routeMatch ? decodeURIComponent(routeMatch[1] || "") : "";
     const warning = document.querySelector(".composer-warning")?.textContent?.trim() || "";
     const input = document.querySelector("#command-input");
-    return Boolean(activeTab && sessionId && sessionId !== initial.sessionId && input && !input.disabled && !warning);
-  }, { sessionId: before?.activeSessionId || "" }, { timeout: 30000 }).catch(() => null);
+    return Boolean(activeTab && sessionId === expectedSessionId && routeSessionId === expectedSessionId && input && !input.disabled && !warning);
+  }, createdSessionId, { timeout: 45000 }).catch(() => null);
   const after = await workbenchSessionSnapshot();
-  const beforeSessionId = before?.activeSessionId || before?.routeSessionId || "";
   const afterSessionId = after?.activeSessionId || after?.routeSessionId || "";
-  const ok = Boolean(afterSessionId && afterSessionId !== beforeSessionId && after?.composerReady);
+  const ok = Boolean(afterSessionId === createdSessionId && after?.routeSessionId === createdSessionId && after?.composerReady);
   if (!ok) {
-    const error = new Error("newSession did not create or select a different workbench session");
-    error.details = { beforeUrl, afterUrl: currentPageUrl(), before, after, pageId, valuesRedacted: true };
+    const error = new Error("newSession did not select the authoritative newly created workbench session");
+    error.details = { beforeUrl, afterUrl: currentPageUrl(), before, after, createdSessionId, pageId, valuesRedacted: true };
     throw error;
   }
   return {
@@ -707,11 +733,20 @@ async function createSessionFromUi() {
     ok,
     before,
     after,
-    sessionId: afterSessionId || null,
+    sessionId: createdSessionId,
+    createSession: { status: createStatus, statusText: createResponse.statusText(), responseParsed: createPayload !== null, responseParseError: createPayloadError, createdSessionId, valuesRedacted: true },
     pageId
   };
 }
 
+function sessionIdFromAgentSessionPayload(payload) {
+  const direct = payload?.sessionId ?? payload?.id ?? payload?.session?.sessionId ?? payload?.session?.id ?? payload?.data?.sessionId ?? payload?.data?.id ?? payload?.data?.session?.sessionId ?? payload?.data?.session?.id;
+  const directText = String(direct || "").trim();
+  if (/^ses_[A-Za-z0-9_-]+$/u.test(directText)) return directText;
+  const match = JSON.stringify(payload ?? "").match(/\bses_[A-Za-z0-9_-]+\b/u);
+  return match ? match[0] : null;
+}
+
 async function workbenchSessionSnapshot(targetPage = page) {
   return targetPage.evaluate(() => {
     const activeTab = document.querySelector(".session-tab[data-active='true'], .session-tab[aria-selected='true']");