[pull] main from danny-avila:main

api/server/controllers/agents/__tests__/request.resumeMetadata.spec.js

@@ -738,6 +738,7 @@ describe('ResumableAgentController resume metadata', () => {
           model: 'gpt-3.5-turbo',
           /** The OWNING replica's seal capability, read by the steer route. */
           preemptCapable: true,
+          steerQuotesCapable: true,
           agent_id: undefined,
           isTemporary: true,
           responseMessageId: expect.stringMatching(/^[0-9a-f-]{36}$/),

api/server/controllers/agents/client.js

@@ -58,6 +58,7 @@ const {
   isSteeringSupported,
   isSteerPreemptSupported,
   buildSteerMedia,
+  collectSteerStampTargets,
   stampSteerPartMedia,
   createActivityLabelWiring,
   createActivityPhaseWiring,
@@ -387,6 +388,9 @@ class AgentClient extends BaseClient {
       ...(item.clientSteerId && { clientSteerId: item.clientSteerId }),
       createdAt: item.createdAt,
       ...(item.files?.length && { files: item.files }),
+      // Persisted separately from the text (mirroring `message.quotes`) so the
+      // UI renders reference blocks and replay re-merges them per turn.
+      ...(item.quotes?.length && { quotes: item.quotes }),
     };
     this.contentParts.push(part);
     this.steerOffsetState.offset += 1;
@@ -1892,22 +1896,33 @@ class AgentClient extends BaseClient {
 
     payload = formattedMessages;
     this.modelBoundSteerFileIdsBySourceMessageId = new Map();
-    if (this.options.resendFiles) {
-      /** Persisted steer parts of past turns replay with their attachments:
-       *  one batched owner-scoped fetch, re-encoded per turn and stamped as a
-       *  transient `media` array (same resend semantics as message files).
-       *  The stamp lands after the loop above finalized its counts, so the
-       *  re-encoded media (minus the text part the steer part already counted)
-       *  is folded into the budget here — large steered attachments must
-       *  shrink the window like any other resent media. */
+    /** Persisted steer parts of past turns replay with their attachments and
+     *  quotes: one batched owner-scoped fetch, re-encoded per turn and
+     *  stamped as a transient `media` array (same resend semantics as
+     *  message files). Runs regardless of `resendFiles` because quote-bearing
+     *  parts must re-merge their excerpts every turn (mirroring
+     *  `prependQuotes` above); file encoding stays gated on the setting via
+     *  the flag. The stamp lands after the loop above finalized its counts,
+     *  so the re-encoded media (minus the text part the steer part already
+     *  counted) is folded into the budget here — large steered attachments
+     *  and quote blocks must shrink the window like any other resent media.
+     *  The synchronous collection keeps steer-free histories on the
+     *  zero-await path to the parallel context kickoff below, and the
+     *  collected targets feed the stamp directly so the history is scanned
+     *  once. */
+    const resendSteerFiles = this.options.resendFiles === true;
+    const steerStampTargets = collectSteerStampTargets(payload, resendSteerFiles);
+    if (steerStampTargets.length > 0) {
       const stamped = await stampSteerPartMedia({
         client: this,
         user: this.options.req?.user,
         payload,
+        targets: steerStampTargets,
         // addPreviousAttachments already fetched steer-part refs in its single
         // per-turn historical-files query — no second round trip.
         docsById: this.authorizedHistoricalFiles,
         getFiles: db.getFiles,
+        resendFiles: resendSteerFiles,
       });
       for (const { sourceMessageId, fileIds } of stamped) {
         if (typeof sourceMessageId !== 'string' || sourceMessageId.length === 0) {
@@ -1927,8 +1942,8 @@ class AgentClient extends BaseClient {
       for (const { index, media, steerText } of stamped) {
         /** Count the FULL stamped content and subtract only the steer body
          *  (already counted inside the assistant message): extracted file
-         *  context prepended into the text part must hit the budget too, or
-         *  large steered documents bypass pruning. */
+         *  context and merged quote blocks prepended into the text part must
+         *  hit the budget too, or large steered documents bypass pruning. */
         const fullTokens = countFormattedMessageTokens({ role: 'user', content: media }, encoding);
         const bodyTokens = steerText
           ? countFormattedMessageTokens(
@@ -1949,6 +1964,25 @@ class AgentClient extends BaseClient {
           memoryFormattedMessages[i] ?? buildMemoryFormattedMessage(orderedMessages[i]),
         );
       }
+      /** The memory copy feeds `processMemory` through the same
+       *  `formatAgentMessages` replay, which reads `part.media`/`part.steer`
+       *  and ignores `part.quotes` — so a steer whose substance lives in its
+       *  quote must be quote-merged here too or memory extraction never sees
+       *  it. Quote merge only (`resendFiles: false`): file media is exactly
+       *  what the memory copy exists to exclude, and text-only stamps touch
+       *  no file fetch or encode. Runs after the fill above so late-built
+       *  copies are stamped too. */
+      const memorySteerTargets = collectSteerStampTargets(memoryPayload, false);
+      if (memorySteerTargets.length > 0) {
+        await stampSteerPartMedia({
+          client: this,
+          user: this.options.req?.user,
+          payload: memoryPayload,
+          targets: memorySteerTargets,
+          getFiles: db.getFiles,
+          resendFiles: false,
+        });
+      }
     }
     this.memoryPayload = hasFileContext ? memoryPayload : null;
     messages = orderedMessages;

api/server/controllers/agents/client.test.js

@@ -3749,6 +3749,60 @@ describe('AgentClient - titleConvo', () => {
       );
     });
 
+    it('quote-merges historical steer parts into the prompt AND the memory copy', async () => {
+      const previousFileContext =
+        'Attached document(s):\n```md\n# "previous.txt"\nPrevious turn file body\n```';
+
+      const result = await client.buildMessages(
+        [
+          {
+            messageId: 'msg-1',
+            parentMessageId: null,
+            sender: 'User',
+            text: 'Summarize.',
+            isCreatedByUser: true,
+            fileContext: previousFileContext,
+          },
+          {
+            messageId: 'msg-2',
+            parentMessageId: 'msg-1',
+            sender: 'Assistant',
+            text: '',
+            isCreatedByUser: false,
+            content: [
+              { type: ContentTypes.TEXT, text: 'working on it' },
+              {
+                type: ContentTypes.STEER,
+                [ContentTypes.STEER]: 'remember this',
+                steerId: 's1',
+                quotes: ['the important fact'],
+              },
+            ],
+          },
+          {
+            messageId: 'msg-3',
+            parentMessageId: 'msg-2',
+            sender: 'User',
+            text: 'Continue.',
+            isCreatedByUser: true,
+          },
+        ],
+        'msg-3',
+        {},
+      );
+
+      const merged = '> the important fact\n\nremember this';
+      const promptSteer = result.prompt[1].content.find((part) => part.type === ContentTypes.STEER);
+      expect(promptSteer.media).toEqual([{ type: ContentTypes.TEXT, text: merged }]);
+      // The memory copy replays through the same formatter, which ignores
+      // `part.quotes` — it needs its own merged stamp or memory extraction
+      // never sees the excerpt.
+      const memorySteer = client.memoryPayload[1].content.find(
+        (part) => part.type === ContentTypes.STEER,
+      );
+      expect(memorySteer.media).toEqual([{ type: ContentTypes.TEXT, text: merged }]);
+    });
+
     it('persists canonical token counts while counting request file context for the prompt', async () => {
       const { countFormattedMessageTokens } = require('@librechat/api');
       const currentFile = makeTextFile('current-file', 'current.txt', 'Current turn file body');

api/server/controllers/agents/request.js

@@ -444,7 +444,7 @@ const ResumableAgentController = async (req, res, next, initializeClient, addTit
   const isRecoveredSteerRequest = recoveredSteerId != null;
   const recoveryUserMessageId = rawOverrideUserMessageId;
   const recoveredSteerPayload = isRecoveredSteerRequest
-    ? buildRecoveredSteerPayload(text, req.body?.files)
+    ? buildRecoveredSteerPayload(text, req.body?.files, req.body?.quotes)
     : undefined;
   /** A recovered steer is handed off as a new ordinary user turn. Edit,
    * regenerate, continue, and arbitrary override-id shapes can reuse an
@@ -1106,6 +1106,10 @@ const ResumableAgentController = async (req, res, next, initializeClient, addTit
         // route may land on a different replica whose own SDK probe would
         // answer for the wrong process during a rolling deploy.
         preemptCapable: isSteerPreemptSupported(),
+        // Same owner-recorded pattern: this build's drain merges queued steer
+        // quotes into the injected turn. Admission on another replica must
+        // not store/acknowledge quotes an older owner would drop.
+        steerQuotesCapable: true,
         // Persist the originating agent so a HITL resume can refuse to rebuild this
         // paused run on a different agent (see resume.js).
         agent_id: endpointOption.agent_id ?? req.body?.agent_id,

api/server/controllers/agents/resume.js

@@ -1222,6 +1222,9 @@ const ResumeAgentController = async (req, res, next, initializeClient, addTitle)
         pendingAction.actionId,
         {
           preemptCapable: isSteerPreemptSupported(),
+          // The handover owner's quote handling replaces the previous
+          // replica's flag, mirroring `preemptCapable` above.
+          steerQuotesCapable: true,
           providerExecutionId,
           providerDrained: true,
           ...(resolvedAskUserQuestion && { resolvedAskUserQuestions }),

client/jest.config.cjs

@@ -25,6 +25,8 @@ module.exports = {
   // },
   moduleNameMapper: {
     '\\.(css)$': 'identity-obj-proxy',
+    /** Mirror the vite resolve.alias so tests parse math with the same tokenizer as production. */
+    '^micromark-extension-math$': 'micromark-extension-llm-math',
     '\\.(jpg|jpeg|png|gif|eot|otf|webp|svg|ttf|woff|woff2|mp4|webm|wav|mp3|m4a|aac|oga)$':
       'jest-file-loader',
     '^test/(.*)$': '<rootDir>/test/$1',

client/package.json

@@ -91,6 +91,8 @@
     "micromark-extension-gfm": "^3.0.0",
     "micromark-extension-llm-math": "^3.1.0",
     "micromark-extension-math": "^3.1.0",
+    "micromark-util-character": "^2.1.0",
+    "micromark-util-symbol": "^2.0.0",
     "monaco-editor": "^0.56.0",
     "qrcode.react": "^4.2.0",
     "rc-input-number": "^7.4.2",
@@ -159,6 +161,7 @@
     "jest-environment-jsdom": "^30.2.0",
     "jest-file-loader": "^1.0.3",
     "jest-junit": "^17.0.0",
+    "micromark-util-types": "^2.0.0",
     "postcss": "^8.5.18",
     "postcss-preset-env": "^11.2.0",
     "tailwindcss": "^3.4.1",

client/src/components/Chat/Input/InFlightSteers.tsx

@@ -16,6 +16,7 @@ import {
 import FilePreviewDialog from '~/components/Chat/Messages/Content/FilePreviewDialog';
 import { supportsGenerationProtocolV2, useArmSteerMutation } from '~/data-provider';
 import { steerOverlayHeightFamily, escalatingSteerFamily } from '~/store/steer';
+import MessageQuotes from '~/components/Chat/Messages/Content/MessageQuotes';
 import MarkdownLite from '~/components/Chat/Messages/Content/MarkdownLite';
 import FileContainer from '~/components/Chat/Input/Files/FileContainer';
 import { useSteerCancel, useSteerReclaim, useLocalize } from '~/hooks';
@@ -482,6 +483,9 @@ const InFlightSteer = memo(function InFlightSteer({
             {localize(preempting ? 'com_ui_steer_in_flight_preempt' : 'com_ui_steer_in_flight')}
           </span>
           <div className="flex min-w-0 flex-col items-start gap-1">
+            {/* Same reference blocks the applied `SteerPart` shows, outside the
+             *  collapse so the excerpts stay visible while a long steer clips. */}
+            <MessageQuotes quotes={steer.quotes} />
             <div
               ref={contentRef}
               id={contentId}

client/src/components/Chat/Input/PendingQuoteChips.tsx

@@ -26,9 +26,11 @@ const CLOSE_DELAY_MS = 120;
  * `autoFocusOnShow` is disabled so opening never pulls focus off the composer;
  * `autoFocusOnHide` still returns focus to the trigger when focus was inside.
  *
- * Reads + writes `pendingQuotesByConvoId` directly; the atom is drained in
- * `useChatFunctions.ask` on submit, so chips disappear once the message is sent
- * (the excerpts then re-render as `MessageQuotes` on the user bubble).
+ * Reads + writes `pendingQuotesByConvoId` directly; the atom is drained on
+ * every submit route — `useChatFunctions.ask` for a fresh send, and the
+ * during-run steer/queue/interrupt paths in `useSteering` — so chips disappear
+ * once the message is sent (the excerpts then re-render as `MessageQuotes` on
+ * the user bubble, or inside the steer bubble for a mid-run injection).
  */
 function PendingQuoteChips({ conversationId }: { conversationId: string }) {
   const localize = useLocalize();

client/src/components/Chat/Input/PendingSteerChips.tsx

@@ -2,7 +2,7 @@ import { memo, useMemo, useRef, useState, useCallback } from 'react';
 import { useAtomValue } from 'jotai';
 import { useRecoilValue } from 'recoil';
 import { useToastContext } from '@librechat/client';
-import { X, Zap, Send, Clock, Pencil, Trash2, Paperclip, RotateCcw } from 'lucide-react';
+import { X, Zap, Send, Clock, Pencil, Trash2, Paperclip, RotateCcw, TextQuote } from 'lucide-react';
 import type { TMessage } from 'librechat-data-provider';
 import type { SteeringControls, QueuedMessageContext } from '~/hooks/Chat/useSteering';
 import type { PendingSteer, QueuedMessage } from '~/store/families';
@@ -24,19 +24,47 @@ import store from '~/store';
 const ROW_CLASS =
   'flex w-full items-center gap-2 rounded-xl border border-border-light bg-surface-secondary px-3 py-2 text-sm text-text-primary';
 
-function AttachmentCount({ count, label }: { count: number; label: string }) {
+function ContextCount({
+  icon,
+  count,
+  label,
+}: {
+  icon: React.ReactNode;
+  count: number;
+  label: string;
+}) {
   if (count === 0) {
     return null;
   }
   return (
     <span className="flex shrink-0 items-center gap-0.5 text-xs text-text-secondary">
-      <Paperclip className="h-3.5 w-3.5" aria-hidden="true" />
+      {icon}
       {count}
       <span className="sr-only">{label}</span>
     </span>
   );
 }
 
+function AttachmentCount({ count, label }: { count: number; label: string }) {
+  return (
+    <ContextCount
+      icon={<Paperclip className="h-3.5 w-3.5" aria-hidden="true" />}
+      count={count}
+      label={label}
+    />
+  );
+}
+
+function QuoteCount({ count, label }: { count: number; label: string }) {
+  return (
+    <ContextCount
+      icon={<TextQuote className="h-3.5 w-3.5" aria-hidden="true" />}
+      count={count}
+      label={label}
+    />
+  );
+}
+
 function QueuedRow({
   message,
   steering,
@@ -61,6 +89,7 @@ function QueuedRow({
   const toggleEntry = useDefaultToggleEntry(steering);
   const interruptToggle = useInterruptToggleEntry();
   const fileCount = message.files?.length ?? 0;
+  const quoteCount = message.quotes?.length ?? 0;
   const isRecovered = message.recoverySteerId != null;
   const actionPendingRef = useRef(false);
   const [actionPending, setActionPending] = useState(false);
@@ -147,6 +176,10 @@ function QueuedRow({
       <span className="min-w-0 flex-1 truncate" title={message.text}>
         {message.text}
       </span>
+      <QuoteCount
+        count={quoteCount}
+        label={localize('com_ui_queued_quote_count', { 0: String(quoteCount) })}
+      />
       <AttachmentCount
         count={fileCount}
         label={localize('com_ui_queued_attachment_count', { 0: String(fileCount) })}
@@ -275,6 +308,10 @@ function FailedSteerRow({
       <span className="min-w-0 flex-1 truncate" title={steer.text}>
         {steer.text}
       </span>
+      <QuoteCount
+        count={steer.quotes?.length ?? 0}
+        label={localize('com_ui_queued_quote_count', { 0: String(steer.quotes?.length ?? 0) })}
+      />
       <span className="shrink-0 text-xs text-red-500">
         {localize(
           steer.deliveryUncertain ? 'com_ui_steer_delivery_unconfirmed' : 'com_ui_steer_failed',

client/src/components/Chat/Input/QuoteButton.tsx

@@ -2,18 +2,15 @@ import { memo, useRef, useState, useEffect, useCallback, useLayoutEffect } from
 import { createPortal } from 'react-dom';
 import { TextQuote } from 'lucide-react';
 import { useSetRecoilState } from 'recoil';
+import { cn, MAX_QUOTE_COUNT } from '~/utils';
 import { mainTextareaId } from '~/common';
 import { useLocalize } from '~/hooks';
-import { cn } from '~/utils';
 import store from '~/store';
 
 /** Only selections fully inside a rendered chat message get the popup. */
 const MESSAGE_SELECTOR = '.message-render';
 /** Max characters captured per excerpt (backend re-caps as defense-in-depth). */
 const MAX_QUOTE_LENGTH = 1500;
-/** Max excerpts queued at once; mirrors the backend `QUOTE_MAX_COUNT` cap so
- *  the composer never shows more quotes than the model actually receives. */
-const MAX_QUOTE_COUNT = 10;
 /** Vertical gap (px) between the selection and the popup. */
 const POPUP_OFFSET = 8;
 /** Keep the popup this far (px) from the viewport edges. */

client/src/components/Chat/Input/__tests__/InFlightSteers.test.tsx

@@ -185,6 +185,19 @@ describe('InFlightSteers', () => {
     expect(screen.queryByTestId('in-flight-steers')).toBeNull();
   });
 
+  it('renders carried quotes as the same reference blocks the applied part shows', () => {
+    renderSteers([
+      {
+        steerId: 's-quoted',
+        text: 'about the selection',
+        status: 'pending',
+        createdAt: 1,
+        quotes: ['the selected excerpt'],
+      },
+    ]);
+    expect(screen.getByTestId('message-quotes')).toHaveTextContent('the selected excerpt');
+  });
+
   it('shows the menu at rest on every pointer, without hover-gating', () => {
     renderSteers([
       { steerId: 's-ack', text: 'waiting on boundary', status: 'pending', createdAt: 1 },