agentgui 1.0.1008 → 1.0.1010

package/AGENTS.md

@@ -1,5 +1,9 @@
 # AgentGUI — Agent Notes
 
+## Mid-thread retry (2026-07-02) — twenty-fifth run
+
+Landed the highest-severity finding deferred from the 24th run (gui-completion #7): retry was hard-gated to the LAST message only — a user unhappy with an earlier assistant reply had no way to redo it. Kit `AgentChat` now offers the retry action on every settled assistant turn, not just `i === lastIdx`; `onRetryMessage(m)` passes the specific message. App `retryTurn(m)` truncates+resends from `m`'s preceding user turn (falls back to the trailing assistant turn when called with no argument, preserving the old "retry last" behavior). **Retrying anything before the trailing turn discards every later turn too** — it now arms the same confirm banner edit-and-resend already uses (generalized with a `kind: 'retry'` tag: "Retry this turn? Retrying will remove the later turns") instead of firing immediately; retrying the actual last turn (no data loss) still executes immediately. Caught and fixed a real bug while generalizing: the old `retryLastTurn()` never cleared `state.chat.resumeSid`/`resumeNote` on truncation, unlike `confirmEditAndResend()` which already did per the "never `--resume` a diverged tail" rule — both paths now share `executeTruncateAndResend()` so this can't drift apart again. Kit pushed `834c81d`, agentgui pushed `18cd971`. Browser-witnessed end-to-end (mid-thread retry click -> confirm banner armed with the correct index -> confirm -> truncated to exactly the retried turn -> resent -> resumeSid cleared), 0 console errors, 28/28 tests. **Note:** this run's Deploy GH Pages CI step flaked twice ("Timeout reached" / "Deployment cancelled") before succeeding on a 3rd rerun — an infra flake in the docs-site deploy environment, not a code regression (`npm publish`, the actual consumer artifact, was green from the first attempt every time).
+
 ## GUI completion sweep (2026-07-02) — twenty-fourth run
 
 `gui-completion` workflow wf_4a731841-832 (33 agents, 17 confirmed). Landed 5 of the smaller-scope findings: Live-tab dashboard state (sort/filter/errors-only) now round-trips through the URL hash (`lsort=`/`lfilter=`/`lerr=1`) and localStorage — previously `lv.filter` wasn't persisted anywhere and the whole dashboard view reverted to defaults on reload. Shortcuts overlay (`?`) now wraps its content in the kit's existing `FocusTrap` component instead of a bare `role=dialog` div — Tab could previously escape into the page behind the overlay. History event rows expose copy even when collapsed (was expanded-only). Settings agents panel gives a not-installed direct-runner agent (claude-code/agy) a "re-check" button, matching the ACP rows' "restart" affordance. **Caught and fixed a real CSRF-guard test gap**: adding the regression test for the upload-PUT route (`scripts/validate-mutations.mjs`), the first draft accidentally included the `Authorization` header — matching every other case in the file but testing the wrong scenario (a stolen-token request, not real cross-site CSRF) — and passed trivially at 200. Corrected to omit auth (the path a real cross-site attacker is actually limited to); the server's existing guard correctly rejects it. **Rule for any CSRF/auth regression test: match the real attacker's capability set exactly — copying a nearby test's header pattern can silently test the wrong threat model.** 12 findings deferred to their own re-scoped PRD row (mid-thread retry/fork, file-management move UX with folder-tree picker, event-level permalinks, rolling context-size indicator — each needs a larger architecture change than a targeted fix). No kit changes this run. Pushed `d3e2d1e`. 28/28 tests, 17/17 validate-mutations, 0 console errors.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentgui",
-  "version": "1.0.1008",
+  "version": "1.0.1010",
   "description": "Multi-agent ACP client with real-time communication",
   "type": "module",
   "main": "electron/main.js",

package/site/app/js/app.js

@@ -1945,11 +1945,12 @@ function chatMain() {
       children: [Spinner({ key: 'trspin', size: 'sm' })] }));
   }
   if (state.chat.confirmingEdit) {
-    banners.push(Alert({ key: 'confedit', kind: 'warn', title: 'Edit this message?',
+    const isRetry = state.chat.confirmingEdit.kind === 'retry';
+    banners.push(Alert({ key: 'confedit', kind: 'warn', title: isRetry ? 'Retry this turn?' : 'Edit this message?',
       children: [
-        h('span', { key: 'cetext' }, 'Editing will remove the later turns - continue? '),
+        h('span', { key: 'cetext' }, (isRetry ? 'Retrying' : 'Editing') + ' will remove the later turns - continue? '),
         Btn({ key: 'ceno', onClick: cancelEditAndResend, children: 'cancel' }),
-        Btn({ key: 'ceyes', danger: true, onClick: confirmEditAndResend, children: 'continue' })] }));
+        Btn({ key: 'ceyes', danger: true, onClick: confirmEditAndResend, children: isRetry ? 'retry' : 'continue' })] }));
   }
   const lastMsg = state.chat.messages.length ? state.chat.messages[state.chat.messages.length - 1] : null;
   const lastErr = lastMsg ? lastMsg.error : null;
@@ -2041,7 +2042,10 @@ function chatMain() {
       // edit-and-resend a user message (two-step: arm a confirm banner first,
       // since the truncation destroys the later turns).
       onCopyMessage: (m) => copyMessageText(m),
-      onRetryMessage: () => retryLastTurn(),
+      // Mid-thread retry (gui-completion #7): the kit now offers retry on
+      // EVERY assistant turn, not just the last - pass the specific message
+      // through so retryTurn truncates from the right position.
+      onRetryMessage: (m) => retryTurn(m),
       confirmEdit: true,
       onArmEdit: (m) => armEditAndResend(m),
       onEditMessage: (m) => armEditAndResend(m),
@@ -2334,28 +2338,56 @@ function chatFollowups() {
   return result;
 }
 // Retry the last assistant turn: drop it and re-send the preceding user message.
-function retryLastTurn() {
-  if (!canSend() && !state.chat.busy) { /* allow when idle */ }
+// Retry a turn: truncates the transcript back to (and resends) the user
+// message that preceded the given assistant message, discarding it and
+// everything after - the same truncate+resend mechanism edit-and-resend
+// uses. `m` is optional (falls back to the trailing assistant turn) so this
+// also serves as "retry last turn" for onRetryMessage on a dangling user
+// message with no reply yet. Mid-thread retry (any assistant turn, not just
+// the last) was a gui-completion #7 finding - the kit's per-message retry
+// action is no longer gated to the last message index for assistant turns.
+function retryTurn(m) {
   if (state.chat.busy) return;
-  // The armed edit confirm refers to indices this retry is about to shift.
-  state.chat.confirmingEdit = null;
   const msgs = state.chat.messages;
-  // Find the trailing assistant turn and the user message before it.
-  let ai = msgs.length - 1;
-  while (ai >= 0 && msgs[ai].role !== 'assistant') ai--;
+  let ai = m ? msgs.indexOf(m) : -1;
+  if (ai < 0) {
+    // Fall back to the trailing assistant turn (no specific message given,
+    // or the given message somehow isn't in the current transcript).
+    ai = msgs.length - 1;
+    while (ai >= 0 && msgs[ai].role !== 'assistant') ai--;
+  }
   if (ai < 0) return;
   let ui = ai - 1;
   while (ui >= 0 && msgs[ui].role !== 'user') ui--;
   if (ui < 0) return;
+  // Retrying anything before the trailing turn silently discards every LATER
+  // turn too - the same destructive-truncation risk edit-and-resend already
+  // confirms for, so mid-thread retry gets the identical arm-then-confirm
+  // banner instead of firing immediately.
+  const discardsLaterTurns = ai < msgs.length - 1;
+  if (discardsLaterTurns) {
+    state.chat.confirmingEdit = { idx: ui, text: msgs[ui].content || '', kind: 'retry' };
+    render();
+    return;
+  }
+  state.chat.confirmingEdit = null;
+  executeTruncateAndResend(ui);
+}
+function executeTruncateAndResend(ui) {
+  const msgs = state.chat.messages;
   const userText = msgs[ui].content || '';
-  // Drop the user+assistant pair (and anything after) and resend the user text.
   state.chat.messages = msgs.slice(0, ui);
   state.chat.totalCost = computeTotalCost();   // discarded turns leave the spend
+  // Never --resume a session whose tail diverged from what the server saw -
+  // a mid-thread retry/edit truncates history the resumed session still has.
+  state.chat.resumeSid = null;
+  state.chat.resumeNote = null;
   state.chat.draft = userText;
   persistChat();
   if (canSend()) sendChat();
   else render();
 }
+function retryLastTurn() { retryTurn(); }
 // Edit-and-resend a user message: two-step. The edit click ARMS a confirmation
 // (truncating destroys the later turns), the banner's continue executes it.
 function armEditAndResend(m) {
@@ -2369,6 +2401,9 @@ function confirmEditAndResend() {
   const ce = state.chat.confirmingEdit;
   if (!ce || state.chat.busy) return;
   state.chat.confirmingEdit = null;
+  // Retry resends the SAME text immediately (no edit step); edit-and-resend
+  // populates the draft so the user can change it before sending.
+  if (ce.kind === 'retry') { executeTruncateAndResend(ce.idx); return; }
   state.chat.messages = state.chat.messages.slice(0, ce.idx);
   state.chat.totalCost = computeTotalCost();   // discarded turns leave the spend
   // Never --resume a session whose tail diverged from what the server saw.