[pull] main from LibreChat-AI:main

Dockerfile

@@ -1,4 +1,4 @@
-# v0.8.8-rc4
+# v0.8.8
 
 # Base node image
 FROM node:24.16.0-alpine AS node

Dockerfile.multi

@@ -1,5 +1,5 @@
 # Dockerfile.multi
-# v0.8.8-rc4
+# v0.8.8
 
 # Set configurable max-old-space-size with default
 ARG NODE_MAX_OLD_SPACE_SIZE=6144

README.md

@@ -51,17 +51,21 @@
   </a>
 </p>
 
-## 🚀 What's New in v0.8.8-rc4
-
-- **Public Agents API docs:** Serve an OpenAPI specification and interactive Swagger UI for inference, events, Agent management, and Skill management.
-- **Attached workspaces (highly experimental):** Isolate workspaces by conversation, load repository instructions, and use bounded queue waits and command timeouts.
-- **Trace Viewer:** Inspect model conversations as ordered steps with roles, Agent identity, tool rounds, previews, and cost.
-- **Skills:** Author or import a Skill and invoke it in the same Agent run, with safer rollback for failed imports.
-- **Agent activity:** Render system events as distinct turns and hold live activity to one stable row.
-- **MCP reliability:** Send per-request headers without hiding tools, coordinate OAuth refresh across replicas, and preserve credentials through provider outages.
-- **Performance:** Stream Markdown incrementally, virtualize model search, and reduce completed Agent message rendering work.
-
-Read the [full v0.8.8-rc4 changelog](https://www.librechat.ai/changelog/v0.8.8-rc4).
+## 🚀 What's New in v0.8.8
+
+- **Agent Management API (beta):** Create, discover, update, and delete Agents; manage Agent files and Skills; and authenticate machine clients through deployment-bound OIDC identities while preserving existing role and Agent access controls. Public OpenAPI and Swagger UI cover inference, events, Agent management, and Skill management.
+- **Attached workspaces (highly experimental):** Select or save a per-Agent default workspace for each managed or personal code worker, then let Agents inspect trees, read and search files, author changes, and run Bash with bounded timeouts. Personal workers support bounded self-service enrollment, readiness status, and per-Agent Git identity.
+- **Background tool controls:** Optionally cancel ordinary background tools, including attached Bash, while keeping detached Subagent execution independent.
+- **Code approval controls:** Choose **Ask**, **Allow**, or **Deny** for file writes and command execution where administrators permit it, including a **Full access** mode for trusted attached environments. File Search and Run Code also honor role grants.
+- **Manual context compaction:** Start a summarize-only turn before the context window fills while preserving recent conversation content according to the deployment's summarization policy.
+- **Context Usage:** Inspect dialogue, retained tool traffic, Agent instructions, cache, cost, and runway pressure without double-counting category subsets.
+- **Unified attachments:** Upload once and let LibreChat route content to the model or extracted text, then provision File Search and Code tools only when needed.
+- **Models:** Added GPT-6 Astra and GPT-6.1 Sol, with Responses API routing and tool-call support.
+- **Agent and chat UI:** Unified tool activity, reasoning, search, and Agent workflows; added one draggable Pinned section for chats and favorites, morphing state icons, high-contrast themes, rich-text message copying, clearer sidebar titles, and refined live phase layouts.
+- **Observability:** Inspect ordered model conversations, tool rounds, and costs in the Trace Viewer; export correlated application logs through OpenTelemetry, configure allowlisted Langfuse trace identity and metadata, tag browser diagnostics with client build IDs, and scope Insights to authorized Agents.
+- **Reliability and security:** Strengthened Agent continuation and checkpoint recovery, Redis liveness detection, DocumentDB coordination, OpenID and MCP OAuth sessions, shared-link throttling, tenant isolation, attachment bounds, and upload error handling.
+
+Read the [full v0.8.8 changelog](https://www.librechat.ai/changelog/v0.8.8).
 
 # ✨ Features
 

api/app/clients/BaseClient.js

@@ -25,6 +25,7 @@ const {
   withBalanceReservations,
   findCheckpointSummaryPart,
   getSummaryPartText,
+  resolveCheckpointMessage,
   runAfterSeed,
   saveTurnConversation,
   seedTurnConversation,
@@ -1419,6 +1420,7 @@ class BaseClient {
    * - The message's 'role' is set to 'system'.
    * - The message's 'text' is set to its 'summary'.
    * - If the message has a 'summaryTokenCount', the message's 'tokenCount' is set to 'summaryTokenCount'.
+   * - A message with a summary content block keeps its content from that block on, for the SDK formatter to promote.
    * The traversal stops at the message with the 'summary' property.
    *
    * Each message object should have an 'id' or 'messageId' property and may have a 'parentMessageId' property.
@@ -1468,35 +1470,13 @@ class BaseClient {
         break;
       }
 
-      let resolved = message;
-      let hasSummary = false;
-      if (summary) {
-        const summaryBlock = findCheckpointSummaryPart(message.content);
-        if (summaryBlock) {
-          const summaryText = getSummaryPartText(summaryBlock);
-          resolved = {
-            ...message,
-            role: 'system',
-            content: [{ type: ContentTypes.TEXT, text: summaryText }],
-            tokenCount: summaryBlock.tokenCount,
-          };
-          hasSummary = true;
-        } else if (message.summary) {
-          resolved = {
-            ...message,
-            role: 'system',
-            content: [{ type: ContentTypes.TEXT, text: message.summary }],
-            tokenCount: message.summaryTokenCount ?? message.tokenCount,
-          };
-          hasSummary = true;
-        }
-      }
-
+      const checkpoint = summary ? resolveCheckpointMessage(message) : null;
+      const resolved = checkpoint ?? message;
       const shouldMap = mapMethod != null && (mapCondition != null ? mapCondition(resolved) : true);
       const processedMessage = shouldMap ? mapMethod(resolved) : resolved;
       orderedMessages.push(processedMessage);
 
-      if (hasSummary) {
+      if (checkpoint) {
         break;
       }
 

api/app/clients/specs/BaseClient.test.js

@@ -548,9 +548,10 @@ describe('BaseClient', () => {
         summary: true,
       });
       expect(result).toHaveLength(2);
-      expect(result[0].role).toBe('system');
-      expect(result[0].content).toEqual([{ type: 'text', text: 'Content block summary' }]);
-      expect(result[0].tokenCount).toBe(42);
+      expect(result[0].role).toBeUndefined();
+      expect(result[0].content).toEqual([
+        { type: 'summary', text: 'Content block summary', tokenCount: 42 },
+      ]);
     });
 
     it('should prefer content block summary over legacy summary field', () => {
@@ -571,8 +572,9 @@ describe('BaseClient', () => {
         summary: true,
       });
       expect(result).toHaveLength(2);
-      expect(result[0].content).toEqual([{ type: 'text', text: 'Content block summary' }]);
-      expect(result[0].tokenCount).toBe(20);
+      expect(result[0].content).toEqual([
+        { type: 'summary', text: 'Content block summary', tokenCount: 20 },
+      ]);
     });
 
     it('should fallback to legacy summary when no content block exists', () => {
@@ -596,6 +598,34 @@ describe('BaseClient', () => {
       expect(result[0].tokenCount).toBe(15);
     });
 
+    it('keeps the parts a response produced after its summary (summary mode)', () => {
+      const trailingText = { type: 'text', text: 'Answer after summarizing' };
+      const messagesWithTrailingParts = [
+        { id: '1', parentMessageId: null, text: 'Message 1' },
+        {
+          id: '2',
+          parentMessageId: '1',
+          text: 'Before summarizing',
+          content: [
+            { type: 'text', text: 'Before summarizing' },
+            { type: 'summary', text: 'Earlier context', tokenCount: 6 },
+            trailingText,
+          ],
+        },
+        { id: '3', parentMessageId: '2', text: 'Message 3' },
+      ];
+      const result = TestClient.constructor.getMessagesForConversation({
+        messages: messagesWithTrailingParts,
+        parentMessageId: '3',
+        summary: true,
+      });
+      expect(result.map((message) => message.id)).toEqual(['2', '3']);
+      expect(result[0].content).toEqual([
+        { type: 'summary', text: 'Earlier context', tokenCount: 6 },
+        trailingText,
+      ]);
+    });
+
     it('should not stop traversal at a failed summary, keeping the prior history', () => {
       /** A summarize round that errored keeps the deltas it streamed, so its
        *  text is a truncated prefix; treating it as the checkpoint would send

api/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@librechat/backend",
-  "version": "v0.8.8-rc4",
+  "version": "v0.8.8",
   "description": "",
   "scripts": {
     "start": "echo 'please run this from the root directory'",
@@ -46,7 +46,7 @@
     "@azure/storage-blob": "^12.30.0",
     "@google/genai": "^2.8.0",
     "@keyv/redis": "5.1.6",
-    "@librechat/agents": "^3.9.7",
+    "@librechat/agents": "^4.0.0",
     "@librechat/api": "*",
     "@librechat/data-schemas": "*",
     "@microsoft/microsoft-graph-client": "^3.0.7",

api/server/controllers/agents/__tests__/client.retainedAnswers.spec.js

@@ -222,6 +222,11 @@ describe('AgentClient retained answers', () => {
     client.user = 'user-123';
     client.processMemory = jest.fn();
     const rows = compactedBranch();
+    /** The stored count covers the whole response, including what preceded its summary. */
+    const storedResponseTokens = 5000;
+    const response = rows.find((row) => row.messageId === 'a2');
+    response.tokenCount = storedResponseTokens;
+    response.content.unshift({ type: ContentTypes.TEXT, text: 'Checking the pipeline.' });
     getMessages.mockResolvedValue(rows);
     /** The real loader: one read of the conversation, then the summary-bounded walk. */
     const cut = await client.loadHistory('convo-123', 'u3');
@@ -239,6 +244,10 @@ describe('AgentClient retained answers', () => {
     expect(text.indexOf(ANSWER_LINE)).toBeLessThan(text.indexOf(LATEST_TEXT));
     expect(text.endsWith(LATEST_TEXT)).toBe(true);
     expect(client.options.agent.additional_instructions ?? '').not.toContain(ANSWER_LINE);
+    expect(text).toContain('Deployed.');
+    expect(text).not.toContain('Checking the pipeline.');
+    expect(tokenCountMap.a2).toBeGreaterThan(0);
+    expect(tokenCountMap.a2).toBeLessThan(storedResponseTokens);
     expect(cut[1].text).toBe(LATEST_TEXT);
     expect(cut[1].content).toBeUndefined();
     expect(counts[prompt.length - 1]).toBe(tokenCountMap.u3);

client/jest.config.cjs

@@ -1,4 +1,4 @@
-/** v0.8.8-rc4 */
+/** v0.8.8 */
 const { maxWorkers } = require('../config/jest.workers.cjs');
 
 module.exports = {

client/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@librechat/frontend",
-  "version": "v0.8.8-rc4",
+  "version": "v0.8.8",
   "description": "",
   "type": "module",
   "scripts": {

client/src/components/Chat/Messages/Content/ToolApproval.tsx

@@ -1,10 +1,12 @@
-import { useEffect, useMemo } from 'react';
+import { useContext, useEffect, useMemo } from 'react';
 import { Button, TextareaAutosize } from '@librechat/client';
-import { Check, X, Pencil, MessageSquare, TriangleAlert } from 'lucide-react';
+import { Check, X, Pencil, MessageSquare, ShieldQuestion, TriangleAlert } from 'lucide-react';
 import type { Agents } from 'librechat-data-provider';
 import type { TranslationKeys } from '~/hooks';
+import { useComposerPresentsApproval } from '~/components/Chat/approval/state';
 import { boundApprovalLabel } from '~/components/Chat/approval/preview';
 import { useApprovalContext, useResumeSubmit } from './ApprovalContext';
+import { ChatContext } from '~/Providers/ChatContext';
 import { useLocalize } from '~/hooks';
 import { cn, logger } from '~/utils';
 
@@ -69,22 +71,28 @@ function seedArgs(args: string | Record<string, unknown> | undefined): string {
  * Renders approve / reject / edit / respond controls for a paused tool call,
  * scoped to the decisions the server allows. Records its decision in the
  * batch {@link useApprovalContext}; the lead card additionally renders the
- * single submit button covering every paused call in the action.
+ * single submit button covering every paused call in the action. While the
+ * composer's review panel is open on the same action, a thread card renders
+ * only the record of the request: the panel owns the decisions and the submit,
+ * and it overlays the tail of the thread where this card sits.
  */
 export default function ToolApproval({
   approval,
   toolCallId,
   args,
-  showSubmit = true,
+  surface = 'thread',
 }: {
   approval: NonNullable<Agents.ToolCall['approval']>;
   toolCallId: string;
   args: string | Record<string, unknown> | undefined;
-  /** The composer owns one batch submit; timeline cards keep the historical lead button. */
-  showSubmit?: boolean;
+  /** The composer panel owns one batch submit; thread cards keep the lead button. */
+  surface?: 'thread' | 'composer';
 }) {
   const localize = useLocalize();
   const { actionId, allowed_decisions: allowedDecisions, description } = approval;
+  const conversationId = useContext(ChatContext)?.conversation?.conversationId;
+  const composerPresents = useComposerPresentsApproval(conversationId, actionId);
+  const deferToComposer = surface === 'thread' && composerPresents;
   const {
     registerToolCall,
     unregisterToolCall,
@@ -209,15 +217,33 @@ export default function ToolApproval({
     return null;
   }
 
+  const descriptionNode = safeDescription != null && safeDescription.length > 0 && (
+    <p className="text-sm text-text-secondary">{safeDescription}</p>
+  );
+
+  if (deferToComposer) {
+    return (
+      <div
+        className="my-2 flex w-full flex-col gap-2 rounded-lg border border-border-light bg-surface-secondary p-3"
+        data-testid="tool-approval"
+        data-tool-call-id={toolCallId}
+      >
+        {descriptionNode}
+        <p className="flex items-center gap-1.5 text-xs text-text-secondary">
+          <ShieldQuestion className="size-4 shrink-0" aria-hidden="true" />
+          {localize('com_ui_approval_review_in_composer')}
+        </p>
+      </div>
+    );
+  }
+
   return (
     <div
       className="my-2 flex w-full flex-col gap-2 rounded-lg border border-border-light bg-surface-secondary p-3"
       data-testid="tool-approval"
       data-tool-call-id={toolCallId}
     >
-      {safeDescription != null && safeDescription.length > 0 && (
-        <p className="text-sm text-text-secondary">{safeDescription}</p>
-      )}
+      {descriptionNode}
       <div className="flex flex-wrap gap-2">
         {allowedDecisions.map((decision) => {
           const Icon = DECISION_ICON[decision];
@@ -281,7 +307,7 @@ export default function ToolApproval({
         />
       )}
 
-      {showSubmit && isLead && (
+      {surface === 'thread' && isLead && (
         <div className="mt-1 flex items-center gap-3">
           <Button
             size="sm"