Brazilian Portuguese translation (#6809)

Co-authored-by: Vitor Alcantara Batista <vitor.alcantara@petrobras.com.br>

.github/ISSUE_TEMPLATE/config.yml

@@ -1,5 +1,5 @@
 blank_issues_enabled: true
 contact_links:
   - name: Jan Discussions
-    url: https://github.com/orgs/menloresearch/discussions/categories/q-a
+    url: https://github.com/orgs/janhq/discussions/categories/q-a
     about: Get help, discuss features & roadmap, and share your projects

.github/workflows/jan-server-web-ci-dev.yml

@@ -12,7 +12,7 @@ jobs:
   build-and-preview:
     runs-on: [ubuntu-24-04-docker]
     env:
-      JAN_API_BASE: "https://api-dev.menlo.ai/v1"
+      MENLO_PLATFORM_BASE_URL: "https://api-dev.jan.ai/v1"
     permissions:
       pull-requests: write
       contents: write
@@ -52,7 +52,7 @@ jobs:
 
       - name: Build docker image
         run: |
-          docker build --build-arg JAN_API_BASE=${{ env.JAN_API_BASE }} -t ${{ steps.vars.outputs.FULL_IMAGE }} .
+          docker build --build-arg MENLO_PLATFORM_BASE_URL=${{ env.MENLO_PLATFORM_BASE_URL }} -t ${{ steps.vars.outputs.FULL_IMAGE }} .
 
       - name: Push docker image
         if: github.event_name == 'push'

.github/workflows/jan-server-web-ci-prod.yml

@@ -13,7 +13,7 @@ jobs:
       deployments: write
       pull-requests: write
     env:
-      JAN_API_BASE: "https://api.menlo.ai/v1"
+      MENLO_PLATFORM_BASE_URL: "https://api.jan.ai/v1"
       GA_MEASUREMENT_ID: "G-YK53MX8M8M"
       CLOUDFLARE_PROJECT_NAME: "jan-server-web"
     steps:
@@ -43,7 +43,7 @@ jobs:
       - name: Install dependencies
         run: make config-yarn && yarn install && yarn build:core && make build-web-app
         env:
-          JAN_API_BASE: ${{ env.JAN_API_BASE }}
+          MENLO_PLATFORM_BASE_URL: ${{ env.MENLO_PLATFORM_BASE_URL }}
           GA_MEASUREMENT_ID: ${{ env.GA_MEASUREMENT_ID }}
 
       - name: Publish to Cloudflare Pages Production

.github/workflows/jan-server-web-ci-stag.yml

@@ -12,7 +12,7 @@ jobs:
   build-and-preview:
     runs-on: [ubuntu-24-04-docker]
     env:
-      JAN_API_BASE: "https://api-stag.menlo.ai/v1"
+      MENLO_PLATFORM_BASE_URL: "https://api-stag.jan.ai/v1"
     permissions:
       pull-requests: write
       contents: write
@@ -52,7 +52,7 @@ jobs:
 
       - name: Build docker image
         run: |
-          docker build --build-arg JAN_API_BASE=${{ env.JAN_API_BASE }} -t ${{ steps.vars.outputs.FULL_IMAGE }} .
+          docker build --build-arg MENLO_PLATFORM_BASE_URL=${{ env.MENLO_PLATFORM_BASE_URL }} -t ${{ steps.vars.outputs.FULL_IMAGE }} .
 
       - name: Push docker image
         if: github.event_name == 'push'

.github/workflows/jan-tauri-build-nightly.yaml

@@ -168,62 +168,62 @@ jobs:
           AWS_DEFAULT_REGION: ${{ secrets.DELTA_AWS_REGION }}
           AWS_EC2_METADATA_DISABLED: 'true'
 
-  noti-discord-nightly-and-update-url-readme:
-    needs:
-      [
-        build-macos,
-        build-windows-x64,
-        build-linux-x64,
-        get-update-version,
-        set-public-provider,
-        sync-temp-to-latest,
-      ]
-    secrets: inherit
-    if: github.event_name == 'schedule'
-    uses: ./.github/workflows/template-noti-discord-and-update-url-readme.yml
-    with:
-      ref: refs/heads/dev
-      build_reason: Nightly
-      push_to_branch: dev
-      new_version: ${{ needs.get-update-version.outputs.new_version }}
+  # noti-discord-nightly-and-update-url-readme:
+  #   needs:
+  #     [
+  #       build-macos,
+  #       build-windows-x64,
+  #       build-linux-x64,
+  #       get-update-version,
+  #       set-public-provider,
+  #       sync-temp-to-latest,
+  #     ]
+  #   secrets: inherit
+  #   if: github.event_name == 'schedule'
+  #   uses: ./.github/workflows/template-noti-discord-and-update-url-readme.yml
+  #   with:
+  #     ref: refs/heads/dev
+  #     build_reason: Nightly
+  #     push_to_branch: dev
+  #     new_version: ${{ needs.get-update-version.outputs.new_version }}
 
-  noti-discord-pre-release-and-update-url-readme:
-    needs:
-      [
-        build-macos,
-        build-windows-x64,
-        build-linux-x64,
-        get-update-version,
-        set-public-provider,
-        sync-temp-to-latest,
-      ]
-    secrets: inherit
-    if: github.event_name == 'push'
-    uses: ./.github/workflows/template-noti-discord-and-update-url-readme.yml
-    with:
-      ref: refs/heads/dev
-      build_reason: Pre-release
-      push_to_branch: dev
-      new_version: ${{ needs.get-update-version.outputs.new_version }}
+  # noti-discord-pre-release-and-update-url-readme:
+  #   needs:
+  #     [
+  #       build-macos,
+  #       build-windows-x64,
+  #       build-linux-x64,
+  #       get-update-version,
+  #       set-public-provider,
+  #       sync-temp-to-latest,
+  #     ]
+  #   secrets: inherit
+  #   if: github.event_name == 'push'
+  #   uses: ./.github/workflows/template-noti-discord-and-update-url-readme.yml
+  #   with:
+  #     ref: refs/heads/dev
+  #     build_reason: Pre-release
+  #     push_to_branch: dev
+  #     new_version: ${{ needs.get-update-version.outputs.new_version }}
 
-  noti-discord-manual-and-update-url-readme:
-    needs:
-      [
-        build-macos,
-        build-windows-x64,
-        build-linux-x64,
-        get-update-version,
-        set-public-provider,
-        sync-temp-to-latest,
-      ]
-    secrets: inherit
-    if: github.event_name == 'workflow_dispatch' && github.event.inputs.public_provider == 'aws-s3'
-    uses: ./.github/workflows/template-noti-discord-and-update-url-readme.yml
-    with:
-      ref: refs/heads/dev
-      build_reason: Manual
-      push_to_branch: dev
-      new_version: ${{ needs.get-update-version.outputs.new_version }}
+  # noti-discord-manual-and-update-url-readme:
+  #   needs:
+  #     [
+  #       build-macos,
+  #       build-windows-x64,
+  #       build-linux-x64,
+  #       get-update-version,
+  #       set-public-provider,
+  #       sync-temp-to-latest,
+  #     ]
+  #   secrets: inherit
+  #   if: github.event_name == 'workflow_dispatch' && github.event.inputs.public_provider == 'aws-s3'
+  #   uses: ./.github/workflows/template-noti-discord-and-update-url-readme.yml
+  #   with:
+  #     ref: refs/heads/dev
+  #     build_reason: Manual
+  #     push_to_branch: dev
+  #     new_version: ${{ needs.get-update-version.outputs.new_version }}
 
   comment-pr-build-url:
     needs:

.github/workflows/jan-tauri-build.yaml

@@ -82,11 +82,11 @@ jobs:
           VERSION=${{ needs.get-update-version.outputs.new_version }}
           PUB_DATE=$(date -u +"%Y-%m-%dT%H:%M:%S.%3NZ")
           LINUX_SIGNATURE="${{ needs.build-linux-x64.outputs.APPIMAGE_SIG }}"
-          LINUX_URL="https://github.com/menloresearch/jan/releases/download/v${{ needs.get-update-version.outputs.new_version }}/${{ needs.build-linux-x64.outputs.APPIMAGE_FILE_NAME }}"
+          LINUX_URL="https://github.com/janhq/jan/releases/download/v${{ needs.get-update-version.outputs.new_version }}/${{ needs.build-linux-x64.outputs.APPIMAGE_FILE_NAME }}"
           WINDOWS_SIGNATURE="${{ needs.build-windows-x64.outputs.WIN_SIG }}"
-          WINDOWS_URL="https://github.com/menloresearch/jan/releases/download/v${{ needs.get-update-version.outputs.new_version }}/${{ needs.build-windows-x64.outputs.FILE_NAME }}"
+          WINDOWS_URL="https://github.com/janhq/jan/releases/download/v${{ needs.get-update-version.outputs.new_version }}/${{ needs.build-windows-x64.outputs.FILE_NAME }}"
           DARWIN_SIGNATURE="${{ needs.build-macos.outputs.MAC_UNIVERSAL_SIG }}"
-          DARWIN_URL="https://github.com/menloresearch/jan/releases/download/v${{ needs.get-update-version.outputs.new_version }}/${{ needs.build-macos.outputs.TAR_NAME }}"
+          DARWIN_URL="https://github.com/janhq/jan/releases/download/v${{ needs.get-update-version.outputs.new_version }}/${{ needs.build-macos.outputs.TAR_NAME }}"
 
           jq --arg version "$VERSION" \
             --arg pub_date "$PUB_DATE" \

.github/workflows/template-get-update-version.yml

@@ -29,7 +29,7 @@ jobs:
             local max_retries=3
             local tag
             while [ $retries -lt $max_retries ]; do
-              tag=$(curl -s https://api.github.com/repos/menloresearch/jan/releases/latest | jq -r .tag_name)
+              tag=$(curl -s https://api.github.com/repos/janhq/jan/releases/latest | jq -r .tag_name)
               if [ -n "$tag" ] && [ "$tag" != "null" ]; then
                 echo $tag
                 return

.github/workflows/template-noti-discord-and-update-url-readme.yml

@@ -50,6 +50,6 @@ jobs:
             - macOS Universal: https://delta.jan.ai/nightly/Jan-nightly_{{ VERSION }}_universal.dmg
             - Linux Deb: https://delta.jan.ai/nightly/Jan-nightly_{{ VERSION }}_amd64.deb
             - Linux AppImage: https://delta.jan.ai/nightly/Jan-nightly_{{ VERSION }}_amd64.AppImage
-            - Github action run: https://github.com/menloresearch/jan/actions/runs/{{ GITHUB_RUN_ID }}
+            - Github action run: https://github.com/janhq/jan/actions/runs/{{ GITHUB_RUN_ID }}
         env:
           DISCORD_WEBHOOK: ${{ secrets.DISCORD_WEBHOOK }}

CONTRIBUTING.md

@@ -143,7 +143,7 @@ jan/
 
 **Option 1: The Easy Way (Make)**
 ```bash
-git clone https://github.com/menloresearch/jan
+git clone https://github.com/janhq/jan
 cd jan
 make dev
 ```
@@ -152,8 +152,8 @@ make dev
 
 ### Reporting Bugs
 
-- **Ensure the bug was not already reported** by searching on GitHub under [Issues](https://github.com/menloresearch/jan/issues)
-- If you're unable to find an open issue addressing the problem, [open a new one](https://github.com/menloresearch/jan/issues/new)
+- **Ensure the bug was not already reported** by searching on GitHub under [Issues](https://github.com/janhq/jan/issues)
+- If you're unable to find an open issue addressing the problem, [open a new one](https://github.com/janhq/jan/issues/new)
 - Include your system specs and error logs - it helps a ton
 
 ### Suggesting Enhancements

Dockerfile

@@ -1,8 +1,8 @@
 # Stage 1: Build stage with Node.js and Yarn v4
 FROM node:20-alpine AS builder
 
-ARG JAN_API_BASE=https://api-dev.jan.ai/v1
-ENV JAN_API_BASE=$JAN_API_BASE
+ARG MENLO_PLATFORM_BASE_URL=https://api-dev.menlo.ai/v1
+ENV MENLO_PLATFORM_BASE_URL=$MENLO_PLATFORM_BASE_URL
 
 # Install build dependencies
 RUN apk add --no-cache \

README.md

@@ -4,10 +4,10 @@
 
 <p align="center">
   <!-- ALL-CONTRIBUTORS-BADGE:START - Do not remove or modify this section -->
-  <img alt="GitHub commit activity" src="https://img.shields.io/github/commit-activity/m/menloresearch/jan"/>
-  <img alt="Github Last Commit" src="https://img.shields.io/github/last-commit/menloresearch/jan"/>
-  <img alt="Github Contributors" src="https://img.shields.io/github/contributors/menloresearch/jan"/>
-  <img alt="GitHub closed issues" src="https://img.shields.io/github/issues-closed/menloresearch/jan"/>
+  <img alt="GitHub commit activity" src="https://img.shields.io/github/commit-activity/m/janhq/jan"/>
+  <img alt="Github Last Commit" src="https://img.shields.io/github/last-commit/janhq/jan"/>
+  <img alt="Github Contributors" src="https://img.shields.io/github/contributors/janhq/jan"/>
+  <img alt="GitHub closed issues" src="https://img.shields.io/github/issues-closed/janhq/jan"/>
   <img alt="Discord" src="https://img.shields.io/discord/1107178041848909847?label=discord"/>
 </p>
 
@@ -15,7 +15,7 @@
   <a href="https://www.jan.ai/docs/desktop">Getting Started</a>
   - <a href="https://discord.gg/Exe46xPMbK">Community</a>
   - <a href="https://jan.ai/changelog">Changelog</a>
-  - <a href="https://github.com/menloresearch/jan/issues">Bug reports</a>
+  - <a href="https://github.com/janhq/jan/issues">Bug reports</a>
 </p>
 
 Jan is bringing the best of open-source AI in an easy-to-use product. Download and run LLMs with **full control** and **privacy**.
@@ -48,7 +48,7 @@ The easiest way to get started is by downloading one of the following versions f
 </table>
 
 
-Download from [jan.ai](https://jan.ai/) or [GitHub Releases](https://github.com/menloresearch/jan/releases).
+Download from [jan.ai](https://jan.ai/) or [GitHub Releases](https://github.com/janhq/jan/releases).
 
 ## Features
 
@@ -73,7 +73,7 @@ For those who enjoy the scenic route:
 ### Run with Make
 
 ```bash
-git clone https://github.com/menloresearch/jan
+git clone https://github.com/janhq/jan
 cd jan
 make dev
 ```
@@ -128,7 +128,7 @@ Contributions welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) for the full spiel
 
 ## Contact
 
-- **Bugs**: [GitHub Issues](https://github.com/menloresearch/jan/issues)
+- **Bugs**: [GitHub Issues](https://github.com/janhq/jan/issues)
 - **Business**: hello@jan.ai
 - **Jobs**: hr@jan.ai
 - **General Discussion**: [Discord](https://discord.gg/FTk2MvZwJH)

autoqa/requirements.txt

@@ -1,7 +1,7 @@
 # Core dependencies
 cua-computer[all]~=0.3.5
 cua-agent[all]~=0.3.0
-cua-agent @ git+https://github.com/menloresearch/cua.git@compute-agent-0.3.0-patch#subdirectory=libs/python/agent
+cua-agent @ git+https://github.com/janhq/cua.git@compute-agent-0.3.0-patch#subdirectory=libs/python/agent
 
 # ReportPortal integration
 reportportal-client~=5.6.5

core/README.md

@@ -13,7 +13,7 @@ import * as core from '@janhq/core'
 
 ## Build an Extension
 
-1. Download an extension template, for example, [https://github.com/menloresearch/extension-template](https://github.com/menloresearch/extension-template).
+1. Download an extension template, for example, [https://github.com/janhq/extension-template](https://github.com/janhq/extension-template).
 
 2. Update the source code:
 

core/package.json

@@ -31,7 +31,7 @@
     "@vitest/coverage-v8": "^2.1.8",
     "@vitest/ui": "^2.1.8",
     "eslint": "8.57.0",
-    "happy-dom": "^15.11.6",
+    "happy-dom": "^20.0.0",
     "pacote": "^21.0.0",
     "react": "19.0.0",
     "request": "^2.88.2",

core/src/browser/extension.ts

@@ -11,6 +11,8 @@ export enum ExtensionTypeEnum {
   HuggingFace = 'huggingFace',
   Engine = 'engine',
   Hardware = 'hardware',
+  RAG = 'rag',
+  VectorDB = 'vectorDB',
 }
 
 export interface ExtensionType {

core/src/browser/extensions/engines/AIEngine.ts

@@ -182,6 +182,7 @@ export interface SessionInfo {
   port: number // llama-server output port (corrected from portid)
   model_id: string //name of the model
   model_path: string // path of the loaded model
+  is_embedding: boolean
   api_key: string
   mmproj_path?: string
 }

core/src/browser/extensions/index.ts

@@ -23,3 +23,8 @@ export { MCPExtension } from './mcp'
  * Base AI Engines.
  */
 export * from './engines'
+
+export { RAGExtension, RAG_INTERNAL_SERVER } from './rag'
+export type { AttachmentInput, IngestAttachmentsResult } from './rag'
+export { VectorDBExtension } from './vector-db'
+export type { SearchMode, VectorDBStatus, VectorChunkInput, VectorSearchResult, AttachmentFileInfo, VectorDBFileInput, VectorDBIngestOptions } from './vector-db'

core/src/browser/extensions/rag.ts

@@ -0,0 +1,36 @@
+import { BaseExtension, ExtensionTypeEnum } from '../extension'
+import type { MCPTool, MCPToolCallResult } from '../../types'
+import type { AttachmentFileInfo } from './vector-db'
+
+export interface AttachmentInput {
+  path: string
+  name?: string
+  type?: string
+  size?: number
+}
+
+export interface IngestAttachmentsResult {
+  filesProcessed: number
+  chunksInserted: number
+  files: AttachmentFileInfo[]
+}
+
+export const RAG_INTERNAL_SERVER = 'rag-internal'
+
+/**
+ * RAG extension base: exposes RAG tools and orchestration API.
+ */
+export abstract class RAGExtension extends BaseExtension {
+  type(): ExtensionTypeEnum | undefined {
+    return ExtensionTypeEnum.RAG
+  }
+
+  abstract getTools(): Promise<MCPTool[]>
+  /**
+   * Lightweight list of tool names for quick routing/lookup.
+   */
+  abstract getToolNames(): Promise<string[]>
+  abstract callTool(toolName: string, args: Record<string, unknown>): Promise<MCPToolCallResult>
+
+  abstract ingestAttachments(threadId: string, files: AttachmentInput[]): Promise<IngestAttachmentsResult>
+}

core/src/browser/extensions/vector-db.ts

@@ -0,0 +1,82 @@
+import { BaseExtension, ExtensionTypeEnum } from '../extension'
+
+export type SearchMode = 'auto' | 'ann' | 'linear'
+
+export interface VectorDBStatus {
+  ann_available: boolean
+}
+
+export interface VectorChunkInput {
+  text: string
+  embedding: number[]
+}
+
+export interface VectorSearchResult {
+  id: string
+  text: string
+  score?: number
+  file_id: string
+  chunk_file_order: number
+}
+
+export interface AttachmentFileInfo {
+  id: string
+  name?: string
+  path?: string
+  type?: string
+  size?: number
+  chunk_count: number
+}
+
+// High-level input types for file ingestion
+export interface VectorDBFileInput {
+  path: string
+  name?: string
+  type?: string
+  size?: number
+}
+
+export interface VectorDBIngestOptions {
+  chunkSize: number
+  chunkOverlap: number
+}
+
+/**
+ * Vector DB extension base: abstraction over local vector storage and search.
+ */
+export abstract class VectorDBExtension extends BaseExtension {
+  type(): ExtensionTypeEnum | undefined {
+    return ExtensionTypeEnum.VectorDB
+  }
+
+  abstract getStatus(): Promise<VectorDBStatus>
+  abstract createCollection(threadId: string, dimension: number): Promise<void>
+  abstract insertChunks(
+    threadId: string,
+    fileId: string,
+    chunks: VectorChunkInput[]
+  ): Promise<void>
+  abstract ingestFile(
+    threadId: string,
+    file: VectorDBFileInput,
+    opts: VectorDBIngestOptions
+  ): Promise<AttachmentFileInfo>
+  abstract searchCollection(
+    threadId: string,
+    query_embedding: number[],
+    limit: number,
+    threshold: number,
+    mode?: SearchMode,
+    fileIds?: string[]
+  ): Promise<VectorSearchResult[]>
+  abstract deleteChunks(threadId: string, ids: string[]): Promise<void>
+  abstract deleteFile(threadId: string, fileId: string): Promise<void>
+  abstract deleteCollection(threadId: string): Promise<void>
+  abstract listAttachments(threadId: string, limit?: number): Promise<AttachmentFileInfo[]>
+  abstract getChunks(
+    threadId: string,
+    fileId: string,
+    startOrder: number,
+    endOrder: number
+  ): Promise<VectorSearchResult[]>
+}

core/src/types/setting/settingComponent.ts

@@ -12,6 +12,8 @@ export type SettingComponentProps = {
   extensionName?: string
   requireModelReload?: boolean
   configType?: ConfigType
+  titleKey?: string
+  descriptionKey?: string
 }
 
 export type ConfigType = 'runtime' | 'setting'

docs/README.md

@@ -18,7 +18,7 @@ We try to **keep routes consistent** to maintain SEO.
 
 ## How to Contribute
 
-Refer to the [Contributing Guide](https://github.com/menloresearch/jan/blob/main/CONTRIBUTING.md) for more comprehensive information on how to contribute to the Jan project.
+Refer to the [Contributing Guide](https://github.com/janhq/jan/blob/main/CONTRIBUTING.md) for more comprehensive information on how to contribute to the Jan project.
 
 ### Pre-requisites and Installation
 

docs/public/openapi/jan.json

@@ -1581,7 +1581,7 @@
           },
           "cover": {
             "type": "string",
-            "example": "https://raw.githubusercontent.com/menloresearch/jan/main/models/trinity-v1.2-7b/cover.png"
+            "example": "https://raw.githubusercontent.com/janhq/jan/main/models/trinity-v1.2-7b/cover.png"
           },
           "engine": {
             "type": "string",

docs/src/components/APIReference/index.tsx

@@ -27,7 +27,7 @@ export const APIReference = () => {
       <ApiReferenceReact
         configuration={{
           spec: {
-            url: 'https://raw.githubusercontent.com/menloresearch/docs/main/public/openapi/jan.json',
+            url: 'https://raw.githubusercontent.com/janhq/docs/main/public/openapi/jan.json',
           },
           theme: 'alternate',
           hideModels: true,

docs/src/components/Changelog/index.tsx

@@ -57,7 +57,7 @@ const Changelog = () => {
           <p className="text-base mt-2 leading-relaxed">
             Latest release updates from the Jan team. Check out our&nbsp;
             <a
-              href="https://github.com/orgs/menloresearch/projects/30"
+              href="https://github.com/orgs/janhq/projects/30"
               className="text-blue-600 dark:text-blue-400 cursor-pointer"
             >
               Roadmap
@@ -150,7 +150,7 @@ const Changelog = () => {
 
       <div className="text-center">
         <Link
-          href="https://github.com/menloresearch/jan/releases"
+          href="https://github.com/janhq/jan/releases"
           target="_blank"
           className="dark:nx-bg-neutral-900 dark:text-white bg-black text-white hover:text-white justify-center dark:border dark:border-neutral-800 flex-shrink-0 px-4 py-3 rounded-xl inline-flex items-center"
         >

docs/src/components/Download/CardDownload.tsx

@@ -72,7 +72,7 @@ export default function CardDownload({ lastRelease }: Props) {
 
           return {
             ...system,
-            href: `https://github.com/menloresearch/jan/releases/download/${lastRelease.tag_name}/${downloadUrl}`,
+            href: `https://github.com/janhq/jan/releases/download/${lastRelease.tag_name}/${downloadUrl}`,
             size: asset ? formatFileSize(asset.size) : undefined,
           }
         })

docs/src/components/DropdownDownload/index.tsx

@@ -139,7 +139,7 @@ const DropdownDownload = ({ lastRelease }: Props) => {
 
           return {
             ...system,
-            href: `https://github.com/menloresearch/jan/releases/download/${lastRelease.tag_name}/${downloadUrl}`,
+            href: `https://github.com/janhq/jan/releases/download/${lastRelease.tag_name}/${downloadUrl}`,
             size: asset ? formatFileSize(asset.size) : undefined,
           }
         })

docs/src/components/Home/BuiltWithLove/index.tsx

@@ -23,7 +23,7 @@ const BuiltWithLove = () => {
         </div>
         <div className="flex flex-col lg:flex-row gap-8 mt-8 items-center justify-center">
           <a
-            href="https://github.com/menloresearch/jan"
+            href="https://github.com/janhq/jan"
             target="_blank"
             className="dark:bg-white bg-black inline-flex w-56 px-4 py-3 rounded-xl cursor-pointer justify-center items-start space-x-4 "
           >

docs/src/components/Home/Hero/index.tsx

@@ -44,7 +44,7 @@ const Hero = () => {
       <div className="mt-10 text-center">
         <div>
           <Link
-            href="https://github.com/menloresearch/jan/releases"
+            href="https://github.com/janhq/jan/releases"
             target="_blank"
             className="hidden lg:inline-block"
           >

docs/src/components/Home/index.tsx

@@ -95,7 +95,7 @@ const Home = () => {
           <div className="container mx-auto relative z-10">
             <div className="flex justify-center items-center mt-14 lg:mt-20 px-4">
               <a
-                href={`https://github.com/menloresearch/jan/releases/tag/${lastVersion}`}
+                href={`https://github.com/janhq/jan/releases/tag/${lastVersion}`}
                 target="_blank"
                 rel="noopener noreferrer"
                 className="bg-black/40 px-3 lg:px-4 rounded-full h-10 inline-flex items-center max-w-full animate-fade-in delay-100"
@@ -270,7 +270,7 @@ const Home = () => {
                   data-delay="600"
                 >
                   <a
-                    href="https://github.com/menloresearch/jan"
+                    href="https://github.com/janhq/jan"
                     target="_blank"
                     rel="noopener noreferrer"
                   >
@@ -387,7 +387,7 @@ const Home = () => {
                       </div>
                       <a
                         className="hidden md:block"
-                        href="https://github.com/menloresearch/jan"
+                        href="https://github.com/janhq/jan"
                         target="_blank"
                         rel="noopener noreferrer"
                       >
@@ -413,7 +413,7 @@ const Home = () => {
                       </p>
                       <a
                         className="md:hidden mt-4 block w-full"
-                        href="https://github.com/menloresearch/jan"
+                        href="https://github.com/janhq/jan"
                         target="_blank"
                         rel="noopener noreferrer"
                       >

docs/src/components/Navbar.tsx

@@ -95,7 +95,7 @@ const Navbar = ({ noScroll }: { noScroll?: boolean }) => {
             })}
             <li>
               <a
-                href="https://github.com/menloresearch/jan/releases/latest"
+                href="https://github.com/janhq/jan/releases/latest"
                 target="_blank"
                 rel="noopener noreferrer"
               >
@@ -141,7 +141,7 @@ const Navbar = ({ noScroll }: { noScroll?: boolean }) => {
                   <FaLinkedinIn className="size-5" />
                 </a>
                 <a
-                  href="https://github.com/menloresearch/jan"
+                  href="https://github.com/janhq/jan"
                   target="_blank"
                   rel="noopener noreferrer"
                   className="rounded-lg flex items-center justify-center"
@@ -156,7 +156,7 @@ const Navbar = ({ noScroll }: { noScroll?: boolean }) => {
         {/* Mobile Download Button and Hamburger */}
         <div className="lg:hidden flex items-center gap-3">
           <a
-            href="https://github.com/menloresearch/jan/releases/latest"
+            href="https://github.com/janhq/jan/releases/latest"
             target="_blank"
             rel="noopener noreferrer"
           >
@@ -278,7 +278,7 @@ const Navbar = ({ noScroll }: { noScroll?: boolean }) => {
                   <FaLinkedinIn className="size-5" />
                 </a>
                 <a
-                  href="https://github.com/menloresearch/jan"
+                  href="https://github.com/janhq/jan"
                   target="_blank"
                   rel="noopener noreferrer"
                   className="text-black rounded-lg flex items-center justify-center"
@@ -296,7 +296,7 @@ const Navbar = ({ noScroll }: { noScroll?: boolean }) => {
                   asChild
                 >
                   <a
-                    href="https://github.com/menloresearch/jan/releases/latest"
+                    href="https://github.com/janhq/jan/releases/latest"
                     target="_blank"
                     rel="noopener noreferrer"
                   >

docs/src/components/ui/dropdown-button.tsx

@@ -120,7 +120,7 @@ export function DropdownButton({
 
           return {
             ...option,
-            href: `https://github.com/menloresearch/jan/releases/download/${lastRelease.tag_name}/${fileName}`,
+            href: `https://github.com/janhq/jan/releases/download/${lastRelease.tag_name}/${fileName}`,
             size: asset ? formatFileSize(asset.size) : 'N/A',
           }
         })

docs/src/pages/api-reference/development.mdx

@@ -18,7 +18,7 @@ description: Development setup, workflow, and contribution guidelines for Jan Se
 
 1. **Clone Repository**
    ```bash
-   git clone https://github.com/menloresearch/jan-server
+   git clone https://github.com/janhq/jan-server
    cd jan-server
    ```
 

docs/src/pages/api-reference/installation.mdx

@@ -19,7 +19,7 @@ Jan Server currently supports minikube for local development. Production Kuberne
 
 1. **Clone the repository**
    ```bash
-   git clone https://github.com/menloresearch/jan-server
+   git clone https://github.com/janhq/jan-server
    cd jan-server
    ```
 

docs/src/pages/changelog/2024-02-10-jan-is-more-stable.mdx

@@ -24,4 +24,4 @@ Fixes 💫
 
 Update your product or download the latest: https://jan.ai
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.5).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.5).

docs/src/pages/changelog/2024-03-06-ui-revamp-settings.mdx

@@ -24,4 +24,4 @@ Jan now supports Mistral's new model Codestral. Thanks [Bartowski](https://huggi
 
 More GGUF models can run in Jan - we rebased to llama.cpp b3012.Big thanks to [ggerganov](https://github.com/ggerganov) 
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.0).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.0).

docs/src/pages/changelog/2024-04-25-llama3-command-r-hugginface.mdx

@@ -28,4 +28,4 @@ Jan now understands LaTeX, allowing users to process and understand complex math
 
 ![Latex](https://catalog.jan.ai/docs/jan_update_latex.gif)
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.4.12).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.4.12).

docs/src/pages/changelog/2024-05-20-llamacpp-upgrade-new-remote-models.mdx

@@ -28,4 +28,4 @@ Users can now connect to OpenAI's new model GPT-4o.
 
 ![GPT4o](https://catalog.jan.ai/docs/jan_v0_4_13_openai_gpt4o.gif)
 
-For more details, see the [GitHub release notes.](https://github.com/menloresearch/jan/releases/tag/v0.4.13)
+For more details, see the [GitHub release notes.](https://github.com/janhq/jan/releases/tag/v0.4.13)

docs/src/pages/changelog/2024-05-28-cohere-aya-23-8b-35b-phi-3-medium.mdx

@@ -16,4 +16,4 @@ More GGUF models can run in Jan - we rebased to llama.cpp b2961.
 
 Huge shoutouts to [ggerganov](https://github.com/ggerganov) and contributors for llama.cpp, and [Bartowski](https://huggingface.co/bartowski) for GGUF models.
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.4.14).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.4.14).

docs/src/pages/changelog/2024-06-21-nvidia-nim-support.mdx

@@ -26,4 +26,4 @@ We've updated to llama.cpp b3088 for better performance - thanks to [GG](https:/
 - Reduced chat font weight (back to normal!)
 - Restored the maximize button
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.1).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.1).

docs/src/pages/changelog/2024-07-15-claude-3-5-support.mdx

@@ -32,4 +32,4 @@ We've restored the tooltip hover functionality, which makes it easier to access
 
 The right-click options for thread settings are now fully operational again. You can now manage your threads with this fix.
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.2).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.2).

docs/src/pages/changelog/2024-09-01-llama3-1-gemma2-support.mdx

@@ -23,4 +23,4 @@ We've been working on stability issues over the last few weeks. Jan is now more
 - Fixed the GPU memory utilization bar
 - Some UX and copy improvements
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.3).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.3).

docs/src/pages/changelog/2024-09-17-improved-cpu-performance.mdx

@@ -32,4 +32,4 @@ Switching between threads used to reset your instruction settings. That’s fixe
 ### Minor UI Tweaks & Bug Fixes
 We’ve also resolved issues with the input slider on the right panel and tackled several smaller bugs to keep everything running smoothly.
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.4).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.4).

docs/src/pages/changelog/2024-10-24-jan-stable.mdx

@@ -23,4 +23,4 @@ Fixes 💫
 
 Update your product or download the latest: https://jan.ai
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.7).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.7).

docs/src/pages/changelog/2024-11-22-jan-bugs.mdx

@@ -22,4 +22,4 @@ Jan v0.5.9 is here: fixing what needed fixing
 
 Update your product or download the latest: https://jan.ai
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.9).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.9).

docs/src/pages/changelog/2024-11.14-jan-supports-qwen-coder.mdx

@@ -22,4 +22,4 @@ and various UI/UX enhancements 💫
 
 Update your product or download the latest: https://jan.ai
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.8).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.8).

docs/src/pages/changelog/2024-12-03-jan-is-faster.mdx

@@ -19,4 +19,4 @@ Jan v0.5.10 is live: Jan is faster, smoother, and more reliable.
 
 Update your product or download the latest: https://jan.ai
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.10).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.10).

docs/src/pages/changelog/2024-12-05-jan-hot-fix-mac.mdx

@@ -23,4 +23,4 @@ Jan v0.5.11 is here - critical issues fixed, Mac installation updated.
 
 Update your product or download the latest: https://jan.ai
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.11).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.11).

docs/src/pages/changelog/2024-12-30-jan-new-privacy.mdx

@@ -25,4 +25,4 @@ Jan v0.5.11 is here - critical issues fixed, Mac installation updated.
 
 Update your product or download the latest: https://jan.ai
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.12).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.12).

docs/src/pages/changelog/2025-01-06-key-issues-resolved.mdx

@@ -20,4 +20,4 @@ import ChangelogHeader from "@/components/Changelog/ChangelogHeader"
 
 Update your product or download the latest: https://jan.ai
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.13).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.13).

docs/src/pages/changelog/2025-01-23-deepseek-r1-jan.mdx

@@ -33,4 +33,4 @@ Llama
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.14).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.14).

docs/src/pages/changelog/2025-02-18-advanced-llama.cpp-settings.mdx

@@ -25,4 +25,4 @@ import ChangelogHeader from "@/components/Changelog/ChangelogHeader"
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.15).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.15).

docs/src/pages/changelog/2025-03-14-jan-security-patch.mdx

@@ -26,4 +26,4 @@ import ChangelogHeader from "@/components/Changelog/ChangelogHeader"
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.16).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.16).

docs/src/pages/changelog/2025-05-14-jan-qwen3-patch.mdx

@@ -20,4 +20,4 @@ import ChangelogHeader from "@/components/Changelog/ChangelogHeader"
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.5.17).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.5.17).

docs/src/pages/changelog/2025-06-19-jan-ui-revamp.mdx

@@ -18,4 +18,4 @@ import ChangelogHeader from "@/components/Changelog/ChangelogHeader"
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.6.1).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.6.1).

docs/src/pages/changelog/2025-06-26-jan-nano-mcp.mdx

@@ -18,4 +18,4 @@ import ChangelogHeader from "@/components/Changelog/ChangelogHeader"
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.6.3).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.6.3).

docs/src/pages/changelog/2025-07-17-responsive-ui.mdx

@@ -23,4 +23,4 @@ new MCP examples.
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For more details, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.6.5).
+For more details, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.6.5).

docs/src/pages/changelog/2025-07-31-llamacpp-tutorials.mdx

@@ -116,4 +116,4 @@ integrations. Stay tuned!
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For the complete list of changes, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.6.6).
+For the complete list of changes, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.6.6).

docs/src/pages/changelog/2025-08-07-gpt-oss.mdx

@@ -89,4 +89,4 @@ We're continuing to optimize performance for large models, expand MCP integratio
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For the complete list of changes, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.6.7).
+For the complete list of changes, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.6.7).

docs/src/pages/changelog/2025-08-14-general-improvs.mdx

@@ -74,4 +74,4 @@ v0.6.8 focuses on stability and real workflows: major llama.cpp hardening, two n
 
 Update your Jan or [download the latest](https://jan.ai/).
 
-For the complete list of changes, see the [GitHub release notes](https://github.com/menloresearch/jan/releases/tag/v0.6.8).
+For the complete list of changes, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.6.8).

docs/src/pages/changelog/2025-10-16-jan-security-update.mdx

@@ -0,0 +1,25 @@
+---
+title: "Jan v0.7.2: Security Update"
+version: 0.7.2
+description: "Jan v0.7.2 updates the happy-dom dependency to v20.0.0 to address a recently disclosed sandbox vulnerability."
+date: 2025-10-16
+---
+
+import ChangelogHeader from "@/components/Changelog/ChangelogHeader"
+import { Callout } from 'nextra/components'
+
+<ChangelogHeader title="Jan v0.7.2" date="2025-10-16" />
+
+## Jan v0.7.2: Security Update (happy-dom v20)
+
+This release focuses on **security and stability improvements**.  
+It updates the `happy-dom` dependency to the latest version to address a recently disclosed vulnerability.
+
+### Security Fix
+- Updated `happy-dom` to **^20.0.0**, preventing untrusted JavaScript executed within HAPPY DOM from accessing process-level functions and executing arbitrary code outside the intended sandbox.
+
+---
+
+Update your Jan or [download the latest version](https://jan.ai/).
+
+For the complete list of changes, see the [GitHub release notes](https://github.com/janhq/jan/releases/tag/v0.7.2).

docs/src/pages/docs/desktop/index.mdx

@@ -41,7 +41,7 @@ Jan is an open-source replacement for ChatGPT:
 
 Jan is a full [product suite](https://en.wikipedia.org/wiki/Software_suite) that offers an alternative to Big AI:
 - [Jan Desktop](/docs/desktop/quickstart): macOS, Windows, and Linux apps with offline mode
-- [Jan Web](https://chat.jan.ai): Jan on browser, a direct alternative to chatgpt.com
+- [Jan Web](https://chat.menlo.ai): Jan on browser, a direct alternative to chatgpt.com
 - Jan Mobile: iOS and Android apps (Coming Soon)
 - [Jan Server](/docs/server): deploy locally, in your cloud, or on-prem
 - [Jan Models](/docs/models): Open-source models optimized for deep research, tool use, and reasoning 

docs/src/pages/docs/desktop/jan-models/jan-nano-128.mdx

@@ -135,5 +135,5 @@ Min-p: 0.0
 
 ## 🤝 Community & Support
 - **Discussions**: [HuggingFace Community](https://huggingface.co/Menlo/Jan-nano-128k/discussions)
-- **Issues**: [GitHub Repository](https://github.com/menloresearch/deep-research/issues)
+- **Issues**: [GitHub Repository](https://github.com/janhq/deep-research/issues)
 - **Discord**: Join our research community for tips and best practices

docs/src/pages/docs/server/architecture.mdx

@@ -9,7 +9,7 @@ Jan Server is a comprehensive self-hosted AI server platform that provides OpenA
 
 Jan Server is a Kubernetes-native platform consisting of multiple microservices that work together to provide a complete AI infrastructure solution. It offers:
 
-![System Architecture Diagram](https://raw.githubusercontent.com/menloresearch/jan-server/main/docs/Architect.png)
+![System Architecture Diagram](https://raw.githubusercontent.com/janhq/jan-server/main/docs/Architect.png)
 
 ### Key Features
 - **OpenAI-Compatible API**: Full compatibility with OpenAI's chat completion API

docs/src/pages/docs/server/development.mdx

@@ -3,7 +3,7 @@ title: Development
 description: Development setup, workflow, and contribution guidelines for Jan Server.
 ---
 ## Core Domain Models
-![Domain Models Diagram](https://github.com/menloresearch/jan-server/raw/main/apps/jan-api-gateway/docs/System_Design.png)
+![Domain Models Diagram](https://github.com/janhq/jan-server/raw/main/apps/jan-api-gateway/docs/System_Design.png)
 ## Development Setup
 
 ### Prerequisites
@@ -42,7 +42,7 @@ description: Development setup, workflow, and contribution guidelines for Jan Se
 
 1. **Clone Repository**
    ```bash
-   git clone https://github.com/menloresearch/jan-server
+   git clone https://github.com/janhq/jan-server
    cd jan-server
    ```
 

docs/src/pages/docs/server/overview.mdx

@@ -40,7 +40,7 @@ Jan Server is a Kubernetes-native platform consisting of multiple microservices
 - **Monitoring & Profiling**: Built-in performance monitoring and health checks
 
 ## System Architecture
-![System Architecture Diagram](https://raw.githubusercontent.com/menloresearch/jan-server/main/docs/Architect.png)
+![System Architecture Diagram](https://raw.githubusercontent.com/janhq/jan-server/main/docs/Architect.png)
 ## Services
 
 ### Jan API Gateway

docs/src/pages/download.mdx

@@ -19,7 +19,7 @@ keywords:
 import Download from "@/components/Download"
 
 export const getStaticProps = async() => {
-  const resRelease = await fetch('https://api.github.com/repos/menloresearch/jan/releases/latest')  
+  const resRelease = await fetch('https://api.github.com/repos/janhq/jan/releases/latest')  
   const release = await resRelease.json()
 
   return {

docs/src/pages/index.mdx

@@ -19,9 +19,9 @@ keywords:
 import Home from "@/components/Home"
 
 export const getStaticProps = async() => {
-  const resReleaseLatest = await fetch('https://api.github.com/repos/menloresearch/jan/releases/latest')  
-  const resRelease = await fetch('https://api.github.com/repos/menloresearch/jan/releases?per_page=500')  
-  const resRepo = await fetch('https://api.github.com/repos/menloresearch/jan')
+  const resReleaseLatest = await fetch('https://api.github.com/repos/janhq/jan/releases/latest')  
+  const resRelease = await fetch('https://api.github.com/repos/janhq/jan/releases?per_page=500')  
+  const resRepo = await fetch('https://api.github.com/repos/janhq/jan')
   const repo = await resRepo.json()
   const latestRelease = await resReleaseLatest.json()
   const release = await resRelease.json()

docs/src/pages/post/benchmarking-nvidia-tensorrt-llm.mdx

@@ -14,12 +14,12 @@ import CTABlog from '@/components/Blog/CTA'
 
 Jan now supports [NVIDIA TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM) in addition to [llama.cpp](https://github.com/ggerganov/llama.cpp), making Jan multi-engine and ultra-fast for users with Nvidia GPUs.
 
-We've been excited for TensorRT-LLM for a while, and [had a lot of fun implementing it](https://github.com/menloresearch/nitro-tensorrt-llm). As part of the process, we've run some benchmarks, to see how TensorRT-LLM fares on consumer hardware (e.g. [4090s](https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/), [3090s](https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/)) we commonly see in the [Jan's hardware community](https://discord.com/channels/1107178041848909847/1201834752206974996).
+We've been excited for TensorRT-LLM for a while, and [had a lot of fun implementing it](https://github.com/janhq/nitro-tensorrt-llm). As part of the process, we've run some benchmarks, to see how TensorRT-LLM fares on consumer hardware (e.g. [4090s](https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/), [3090s](https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/)) we commonly see in the [Jan's hardware community](https://discord.com/channels/1107178041848909847/1201834752206974996).
 
 <Callout type="info" >
   **Give it a try!** Jan's TensorRT-LLM extension is available in Jan v0.4.9. We precompiled some TensorRT-LLM models for you to try: `Mistral 7b`, `TinyLlama-1.1b`, `TinyJensen-1.1b` 😂
 
-  Bugs or feedback? Let us know on [GitHub](https://github.com/menloresearch/jan) or via [Discord](https://discord.com/channels/1107178041848909847/1201832734704795688).
+  Bugs or feedback? Let us know on [GitHub](https://github.com/janhq/jan) or via [Discord](https://discord.com/channels/1107178041848909847/1201832734704795688).
 </Callout>
 
 <Callout type="info" >

docs/src/pages/post/deepresearch.mdx

@@ -70,34 +70,34 @@ brief survey of how other players approach deep research:
 | Kimi | Interactive synthesis | 50–100 | 30–60+ | PDF, Interactive website | Free |
 
 In our testing, we used the following prompt to assess the quality of the generated report by
-the providers above. You can refer to the reports generated [here](https://github.com/menloresearch/prompt-experiments).
+the providers above. You can refer to the reports generated [here](https://github.com/janhq/prompt-experiments).
 
 ```
 Generate a comprehensive report about the state of AI in the past week. Include all
 new model releases and notable architectural improvements from a variety of sources.
 ```
 
-[Google's generated report](https://github.com/menloresearch/prompt-experiments/blob/main/Gemini%202.5%20Flash%20Report.pdf) was the most verbose, with a whopping 23 pages that reads
+[Google's generated report](https://github.com/janhq/prompt-experiments/blob/main/Gemini%202.5%20Flash%20Report.pdf) was the most verbose, with a whopping 23 pages that reads
 like a professional intelligence briefing. It opens with an executive summary,
 systematically categorizes developments, and provides forward-looking strategic
 insights—connecting OpenAI's open-weight release to broader democratization trends
 and linking infrastructure investments to competitive positioning.
 
-[OpenAI](https://github.com/menloresearch/prompt-experiments/blob/main/OpenAI%20Deep%20Research.pdf) produced the most citation-heavy output with 134 references throughout 10 pages
+[OpenAI](https://github.com/janhq/prompt-experiments/blob/main/OpenAI%20Deep%20Research.pdf) produced the most citation-heavy output with 134 references throughout 10 pages
 (albeit most of them being from the same source).
 
-[Perplexity](https://github.com/menloresearch/prompt-experiments/blob/main/Perplexity%20Deep%20Research.pdf) delivered the most actionable 6-page report that maximizes information
+[Perplexity](https://github.com/janhq/prompt-experiments/blob/main/Perplexity%20Deep%20Research.pdf) delivered the most actionable 6-page report that maximizes information
 density while maintaining scannability. Despite being the shortest, it captures all
 major developments with sufficient context for decision-making.
 
-[Claude](https://github.com/menloresearch/prompt-experiments/blob/main/Claude%20Deep%20Research.pdf) produced a comprehensive analysis that interestingly ignored the time constraint,
+[Claude](https://github.com/janhq/prompt-experiments/blob/main/Claude%20Deep%20Research.pdf) produced a comprehensive analysis that interestingly ignored the time constraint,
 covering an 8-month period from January-August 2025 instead of the requested week (Jul 31-Aug
 7th 2025). Rather than cataloging recent events, Claude traced the evolution of trends over months.
 
-[Grok](https://github.com/menloresearch/prompt-experiments/blob/main/Grok%203%20Deep%20Research.pdf) produced a well-structured but relatively shallow 5-page academic-style report that
+[Grok](https://github.com/janhq/prompt-experiments/blob/main/Grok%203%20Deep%20Research.pdf) produced a well-structured but relatively shallow 5-page academic-style report that
 read more like an event catalog than strategic analysis.
 
-[Kimi](https://github.com/menloresearch/prompt-experiments/blob/main/Kimi%20AI%20Deep%20Research.pdf) produced a comprehensive 13-page report with systematic organization covering industry developments, research breakthroughs, and policy changes, but notably lacks proper citations throughout most of the content despite claiming to use 50-100 sources.
+[Kimi](https://github.com/janhq/prompt-experiments/blob/main/Kimi%20AI%20Deep%20Research.pdf) produced a comprehensive 13-page report with systematic organization covering industry developments, research breakthroughs, and policy changes, but notably lacks proper citations throughout most of the content despite claiming to use 50-100 sources.
 
 ### Understanding Search Strategies
 

docs/src/pages/post/rag-is-not-enough.mdx

@@ -13,7 +13,7 @@ import CTABlog from '@/components/Blog/CTA'
 
 ## Abstract
 
-We present a straightforward approach to customizing small, open-source models using fine-tuning and RAG that outperforms GPT-3.5 for specialized use cases. With it, we achieved superior Q&A results of [technical documentation](https://nitro.jan.ai/docs) for a small codebase [codebase](https://github.com/menloresearch/nitro).
+We present a straightforward approach to customizing small, open-source models using fine-tuning and RAG that outperforms GPT-3.5 for specialized use cases. With it, we achieved superior Q&A results of [technical documentation](https://nitro.jan.ai/docs) for a small codebase [codebase](https://github.com/janhq/nitro).
 
 In short, (1) extending a general foundation model like [Mistral](https://huggingface.co/mistralai/Mistral-7B-v0.1) with strong math and coding, and (2) training it over a high-quality, synthetic dataset generated from the intended corpus, and (3) adding RAG capabilities, can lead to significant accuracy improvements.
 
@@ -93,11 +93,11 @@ This final model can be found [here on Huggingface](https://huggingface.co/jan-h
 
 As an additional step, we also added [Retrieval Augmented Generation (RAG)](https://blogs.nvidia.com/blog/what-is-retrieval-augmented-generation/) as an experiment parameter.
 
-A simple RAG setup was done using **[Llamaindex](https://www.llamaindex.ai/)** and the **[bge-en-base-v1.5 embedding](https://huggingface.co/BAAI/bge-base-en-v1.5)** model for efficient documentation retrieval and question-answering. You can find the RAG implementation [here](https://github.com/menloresearch/open-foundry/blob/main/rag-is-not-enough/rag/nitro_rag.ipynb).
+A simple RAG setup was done using **[Llamaindex](https://www.llamaindex.ai/)** and the **[bge-en-base-v1.5 embedding](https://huggingface.co/BAAI/bge-base-en-v1.5)** model for efficient documentation retrieval and question-answering. You can find the RAG implementation [here](https://github.com/janhq/open-foundry/blob/main/rag-is-not-enough/rag/nitro_rag.ipynb).
 
 ## Benchmarking the Results
 
-We curated a new set of [50 multiple-choice questions](https://github.com/menloresearch/open-foundry/blob/main/rag-is-not-enough/rag/mcq_nitro.csv) (MCQ) based on the Nitro docs. The questions had varying levels of difficulty and had trick components that challenged the model's ability to discern misleading information.
+We curated a new set of [50 multiple-choice questions](https://github.com/janhq/open-foundry/blob/main/rag-is-not-enough/rag/mcq_nitro.csv) (MCQ) based on the Nitro docs. The questions had varying levels of difficulty and had trick components that challenged the model's ability to discern misleading information.
 
 ![image](https://hackmd.io/_uploads/By9vaE1Ta.png)
 
@@ -121,7 +121,7 @@ We conclude that this combination of model merging + finetuning + RAG yields pro
 
 Anecdotally, we’ve had some success using this model in practice to onboard new team members to the Nitro codebase.
 
-A full research report with more statistics can be found [here](https://github.com/menloresearch/open-foundry/blob/main/rag-is-not-enough/README.md).
+A full research report with more statistics can be found [here](https://github.com/janhq/open-foundry/blob/main/rag-is-not-enough/README.md).
 
 # References
 

docs/src/pages/post/run-gpt-oss-locally.mdx

@@ -203,7 +203,7 @@ When to choose ChatGPT Plus instead:
 
 Ready to try gpt-oss?
 - Download Jan: [https://jan.ai/](https://jan.ai/)
-- View source code: [https://github.com/menloresearch/jan](https://github.com/menloresearch/jan)
+- View source code: [https://github.com/janhq/jan](https://github.com/janhq/jan)
 - Need help? Check our [local AI guide](/post/run-ai-models-locally) for beginners
 
 <CTABlog />

docs/src/pages/support.mdx

@@ -4,7 +4,7 @@ title: Support - Jan
 
 # Support
 
-- Bugs & requests: file a GitHub ticket [here](https://github.com/menloresearch/jan/issues)
+- Bugs & requests: file a GitHub ticket [here](https://github.com/janhq/jan/issues)
 - For discussion: join our Discord [here](https://discord.gg/FTk2MvZwJH)
 - For business inquiries: email hello@jan.ai
 - For jobs: please email hr@jan.ai

docs/theme.config.tsx

@@ -31,7 +31,7 @@ const config: DocsThemeConfig = {
       </div>
     </span>
   ),
-  docsRepositoryBase: 'https://github.com/menloresearch/jan/tree/dev/docs',
+  docsRepositoryBase: 'https://github.com/janhq/jan/tree/dev/docs',
   feedback: {
     content: 'Question? Give us feedback →',
     labels: 'feedback',

extensions-web/src/conversational-web/api.ts

@@ -16,7 +16,7 @@ import {
   ListConversationItemsResponse
 } from './types'
 
-declare const JAN_API_BASE: string
+declare const MENLO_PLATFORM_BASE_URL: string
 
 export class RemoteApi {
   private authService: JanAuthService
@@ -28,7 +28,7 @@ export class RemoteApi {
   async createConversation(
     data: Conversation
   ): Promise<ConversationResponse> {
-    const url = `${JAN_API_BASE}${CONVERSATION_API_ROUTES.CONVERSATIONS}`
+    const url = `${MENLO_PLATFORM_BASE_URL}${CONVERSATION_API_ROUTES.CONVERSATIONS}`
 
     return this.authService.makeAuthenticatedRequest<ConversationResponse>(
       url,
@@ -43,12 +43,12 @@ export class RemoteApi {
     conversationId: string,
     data: Conversation
   ): Promise<ConversationResponse> {
-    const url = `${JAN_API_BASE}${CONVERSATION_API_ROUTES.CONVERSATION_BY_ID(conversationId)}`
+    const url = `${MENLO_PLATFORM_BASE_URL}${CONVERSATION_API_ROUTES.CONVERSATION_BY_ID(conversationId)}`
 
     return this.authService.makeAuthenticatedRequest<ConversationResponse>(
       url,
       {
-        method: 'PATCH',
+        method: 'POST',
         body: JSON.stringify(data),
       }
     )
@@ -70,7 +70,7 @@ export class RemoteApi {
     }
 
     const queryString = queryParams.toString()
-    const url = `${JAN_API_BASE}${CONVERSATION_API_ROUTES.CONVERSATIONS}${queryString ? `?${queryString}` : ''}`
+    const url = `${MENLO_PLATFORM_BASE_URL}${CONVERSATION_API_ROUTES.CONVERSATIONS}${queryString ? `?${queryString}` : ''}`
 
     return this.authService.makeAuthenticatedRequest<ListConversationsResponse>(
       url,
@@ -114,7 +114,7 @@ export class RemoteApi {
   }
 
   async deleteConversation(conversationId: string): Promise<void> {
-    const url = `${JAN_API_BASE}${CONVERSATION_API_ROUTES.CONVERSATION_BY_ID(conversationId)}`
+    const url = `${MENLO_PLATFORM_BASE_URL}${CONVERSATION_API_ROUTES.CONVERSATION_BY_ID(conversationId)}`
 
     await this.authService.makeAuthenticatedRequest(
       url,
@@ -141,7 +141,7 @@ export class RemoteApi {
     }
 
     const queryString = queryParams.toString()
-    const url = `${JAN_API_BASE}${CONVERSATION_API_ROUTES.CONVERSATION_ITEMS(conversationId)}${queryString ? `?${queryString}` : ''}`
+    const url = `${MENLO_PLATFORM_BASE_URL}${CONVERSATION_API_ROUTES.CONVERSATION_ITEMS(conversationId)}${queryString ? `?${queryString}` : ''}`
 
     return this.authService.makeAuthenticatedRequest<ListConversationItemsResponse>(
       url,

extensions-web/src/conversational-web/types.ts

@@ -31,7 +31,7 @@ export interface ConversationResponse {
   id: string
   object: 'conversation'
   title?: string
-  created_at: number
+  created_at: number | string
   metadata: ConversationMetadata
 }
 
@@ -50,6 +50,7 @@ export interface ConversationItemAnnotation {
 }
 
 export interface ConversationItemContent {
+  type?: string
   file?: {
     file_id?: string
     mime_type?: string
@@ -62,23 +63,50 @@ export interface ConversationItemContent {
     file_id?: string
     url?: string
   }
+  image_file?: {
+    file_id?: string
+    mime_type?: string
+  }
   input_text?: string
   output_text?: {
     annotations?: ConversationItemAnnotation[]
     text?: string
   }
-  reasoning_content?: string
   text?: {
     value?: string
+    text?: string
   }
-  type?: string
+  reasoning_content?: string
+  tool_calls?: Array<{
+    id?: string
+    type?: string
+    function?: {
+      name?: string
+      arguments?: string
+    }
+  }>
+  tool_call_id?: string
+  tool_result?: {
+    content?: Array<{
+      type?: string
+      text?: string
+      output_text?: {
+        text?: string
+      }
+    }>
+    output_text?: {
+      text?: string
+    }
+  }
+  text_result?: string
 }
 
 export interface ConversationItem {
   content?: ConversationItemContent[]
-  created_at: number
+  created_at: number | string
   id: string
   object: string
+  metadata?: Record<string, unknown>
   role: string
   status?: string
   type?: string

extensions-web/src/conversational-web/utils.ts

@@ -1,5 +1,5 @@
 import { Thread, ThreadAssistantInfo, ThreadMessage, ContentType } from '@janhq/core'
-import { Conversation, ConversationResponse, ConversationItem } from './types'
+import { Conversation, ConversationResponse, ConversationItem, ConversationItemContent, ConversationMetadata } from './types'
 import { DEFAULT_ASSISTANT } from './const'
 
 export class ObjectParser {
@@ -7,7 +7,7 @@ export class ObjectParser {
     const modelName = thread.assistants?.[0]?.model?.id || undefined
     const modelProvider = thread.assistants?.[0]?.model?.engine || undefined
     const isFavorite = thread.metadata?.is_favorite?.toString() || 'false'
-    let metadata = {}
+    let metadata: ConversationMetadata = {}
     if (modelName && modelProvider) {
       metadata = {
         model_id: modelName,
@@ -23,15 +23,14 @@ export class ObjectParser {
 
   static conversationToThread(conversation: ConversationResponse): Thread {
     const assistants: ThreadAssistantInfo[] = []
-    if (
-      conversation.metadata?.model_id &&
-      conversation.metadata?.model_provider
-    ) {
+    const metadata: ConversationMetadata = conversation.metadata || {}
+
+    if (metadata.model_id && metadata.model_provider) {
       assistants.push({
         ...DEFAULT_ASSISTANT,
         model: {
-          id: conversation.metadata.model_id,
-          engine: conversation.metadata.model_provider,
+          id: metadata.model_id,
+          engine: metadata.model_provider,
         },
       })
     } else {
@@ -44,16 +43,18 @@ export class ObjectParser {
       })
     }
 
-    const isFavorite = conversation.metadata?.is_favorite === 'true'
+    const isFavorite = metadata.is_favorite === 'true'
+    const createdAtMs = parseTimestamp(conversation.created_at)
+
     return {
       id: conversation.id,
       title: conversation.title || '',
       assistants,
-      created: conversation.created_at,
-      updated: conversation.created_at,
+      created: createdAtMs,
+      updated: createdAtMs,
       model: {
-        id: conversation.metadata.model_id,
-        provider: conversation.metadata.model_provider,
+        id: metadata.model_id,
+        provider: metadata.model_provider,
       },
       isFavorite,
       metadata: { is_favorite: isFavorite },
@@ -65,74 +66,70 @@ export class ObjectParser {
     threadId: string
   ): ThreadMessage {
     // Extract text content and metadata from the item
-    let textContent = ''
-    let reasoningContent = ''
+    const textSegments: string[] = []
+    const reasoningSegments: string[] = []
     const imageUrls: string[] = []
     let toolCalls: any[] = []
-    let finishReason = ''
 
     if (item.content && item.content.length > 0) {
       for (const content of item.content) {
-        // Handle text content
-        if (content.text?.value) {
-          textContent = content.text.value
-        }
-        // Handle output_text for assistant messages
-        if (content.output_text?.text) {
-          textContent = content.output_text.text
-        }
-        // Handle reasoning content
-        if (content.reasoning_content) {
-          reasoningContent = content.reasoning_content
-        }
-        // Handle image content
-        if (content.image?.url) {
-          imageUrls.push(content.image.url)
-        }
-        // Extract finish_reason
-        if (content.finish_reason) {
-          finishReason = content.finish_reason
-        }
-      }
-    }
-
-    // Handle tool calls parsing for assistant messages
-    if (item.role === 'assistant' && finishReason === 'tool_calls') {
-      try {
-        // Tool calls are embedded as JSON string in textContent
-        const toolCallMatch = textContent.match(/\[.*\]/)
-        if (toolCallMatch) {
-          const toolCallsData = JSON.parse(toolCallMatch[0])
-          toolCalls = toolCallsData.map((toolCall: any) => ({
-            tool: {
-              id: toolCall.id || 'unknown',
-              function: {
-                name: toolCall.function?.name || 'unknown',
-                arguments: toolCall.function?.arguments || '{}'
-              },
-              type: toolCall.type || 'function'
-            },
-            response: {
-              error: '',
-              content: []
-            },
-            state: 'ready'
-          }))
-          // Remove tool calls JSON from text content, keep only reasoning
-          textContent = ''
-        }
-      } catch (error) {
-        console.error('Failed to parse tool calls:', error)
+        extractContentByType(content, {
+          onText: (value) => {
+            if (value) {
+              textSegments.push(value)
+            }
+          },
+          onReasoning: (value) => {
+            if (value) {
+              reasoningSegments.push(value)
+            }
+          },
+          onImage: (url) => {
+            if (url) {
+              imageUrls.push(url)
+            }
+          },
+          onToolCalls: (calls) => {
+            toolCalls = calls.map((toolCall) => {
+              const callId = toolCall.id || 'unknown'
+              const rawArgs = toolCall.function?.arguments
+              const normalizedArgs =
+                typeof rawArgs === 'string'
+                  ? rawArgs
+                  : JSON.stringify(rawArgs ?? {})
+              return {
+                id: callId,
+                tool_call_id: callId,
+                tool: {
+                  id: callId,
+                  function: {
+                    name: toolCall.function?.name || 'unknown',
+                    arguments: normalizedArgs,
+                  },
+                  type: toolCall.type || 'function',
+                },
+                response: {
+                  error: '',
+                  content: [],
+                },
+                state: 'pending',
+              }
+            })
+          },
+        })
       }
     }
 
     // Format final content with reasoning if present
     let finalTextValue = ''
-    if (reasoningContent) {
-      finalTextValue = `<think>${reasoningContent}</think>`
+    if (reasoningSegments.length > 0) {
+      finalTextValue += `<think>${reasoningSegments.join('\n')}</think>`
     }
-    if (textContent) {
-      finalTextValue += textContent
+    if (textSegments.length > 0) {
+      if (finalTextValue) {
+        finalTextValue += '\n'
+      }
+      finalTextValue += textSegments.join('\n')
     }
 
     // Build content array for ThreadMessage
@@ -157,22 +154,26 @@ export class ObjectParser {
     }
 
     // Build metadata
-    const metadata: any = {}
+    const metadata: any = { ...(item.metadata || {}) }
     if (toolCalls.length > 0) {
       metadata.tool_calls = toolCalls
     }
 
+    const createdAtMs = parseTimestamp(item.created_at)
+
     // Map status from server format to frontend format
     const mappedStatus = item.status === 'completed' ? 'ready' : item.status || 'ready'
 
+    const role = item.role === 'user' || item.role === 'assistant' ? item.role : 'assistant'
+
     return {
       type: 'text',
       id: item.id,
       object: 'thread.message',
       thread_id: threadId,
-      role: item.role as 'user' | 'assistant',
+      role,
       content: messageContent,
-      created_at: item.created_at * 1000, // Convert to milliseconds
+      created_at: createdAtMs,
       completed_at: 0,
       status: mappedStatus,
       metadata,
@@ -201,25 +202,46 @@ export const combineConversationItemsToMessages = (
 ): ThreadMessage[] => {
   const messages: ThreadMessage[] = []
   const toolResponseMap = new Map<string, any>()
+  const sortedItems = [...items].sort(
+    (a, b) => parseTimestamp(a.created_at) - parseTimestamp(b.created_at)
+  )
 
   // First pass: collect tool responses
-  for (const item of items) {
+  for (const item of sortedItems) {
     if (item.role === 'tool') {
-      const toolContent = item.content?.[0]?.text?.value || ''
-      toolResponseMap.set(item.id, {
-        error: '',
-        content: [
-          {
-            type: 'text',
-            text: toolContent
-          }
-        ]
-      })
+      for (const content of item.content ?? []) {
+        const toolCallId = content.tool_call_id || item.id
+        const toolResultText =
+          content.tool_result?.output_text?.text ||
+          (Array.isArray(content.tool_result?.content)
+            ? content.tool_result?.content
+                ?.map((entry) => entry.text || entry.output_text?.text)
+                .filter((text): text is string => Boolean(text))
+                .join('\n')
+            : undefined)
+        const toolContent =
+          content.text?.text ||
+          content.text?.value ||
+          content.output_text?.text ||
+          content.input_text ||
+          content.text_result ||
+          toolResultText ||
+          ''
+        toolResponseMap.set(toolCallId, {
+          error: '',
+          content: [
+            {
+              type: 'text',
+              text: toolContent,
+            },
+          ],
+        })
+      }
     }
   }
 
   // Second pass: build messages and merge tool responses
-  for (const item of items) {
+  for (const item of sortedItems) {
     // Skip tool messages as they will be merged into assistant messages
     if (item.role === 'tool') {
       continue
@@ -228,14 +250,35 @@ export const combineConversationItemsToMessages = (
     const message = ObjectParser.conversationItemToThreadMessage(item, threadId)
 
     // If this is an assistant message with tool calls, merge tool responses
-    if (message.role === 'assistant' && message.metadata?.tool_calls && Array.isArray(message.metadata.tool_calls)) {
+    if (
+      message.role === 'assistant' &&
+      message.metadata?.tool_calls &&
+      Array.isArray(message.metadata.tool_calls)
+    ) {
       const toolCalls = message.metadata.tool_calls as any[]
-      let toolResponseIndex = 0
 
-      for (const [responseId, responseData] of toolResponseMap.entries()) {
-        if (toolResponseIndex < toolCalls.length) {
-          toolCalls[toolResponseIndex].response = responseData
-          toolResponseIndex++
+      for (const toolCall of toolCalls) {
+        const callId = toolCall.tool_call_id || toolCall.id || toolCall.tool?.id
+        let responseKey: string | undefined
+        let response: any = null
+
+        if (callId && toolResponseMap.has(callId)) {
+          responseKey = callId
+          response = toolResponseMap.get(callId)
+        } else {
+          const iterator = toolResponseMap.entries().next()
+          if (!iterator.done) {
+            responseKey = iterator.value[0]
+            response = iterator.value[1]
+          }
+        }
+
+        if (response) {
+          toolCall.response = response
+          toolCall.state = 'succeeded'
+          if (responseKey) {
+            toolResponseMap.delete(responseKey)
+          }
         }
       }
     }
@@ -245,3 +288,79 @@ export const combineConversationItemsToMessages = (
 
   return messages
 }
+
+const parseTimestamp = (value: number | string | undefined): number => {
+  if (typeof value === 'number') {
+    // Distinguish between seconds and milliseconds
+    return value > 1e12 ? value : value * 1000
+  }
+  if (typeof value === 'string') {
+    const parsed = Date.parse(value)
+    return Number.isNaN(parsed) ? Date.now() : parsed
+  }
+  return Date.now()
+}
+
+const extractContentByType = (
+  content: ConversationItemContent,
+  handlers: {
+    onText: (value: string) => void
+    onReasoning: (value: string) => void
+    onImage: (url: string) => void
+    onToolCalls: (calls: NonNullable<ConversationItemContent['tool_calls']>) => void
+  }
+) => {
+  const type = content.type || ''
+
+  switch (type) {
+    case 'input_text':
+      handlers.onText(content.input_text || '')
+      break
+    case 'text':
+      handlers.onText(content.text?.text || content.text?.value || '')
+      break
+    case 'output_text':
+      handlers.onText(content.output_text?.text || '')
+      break
+    case 'reasoning_content':
+      handlers.onReasoning(content.reasoning_content || '')
+      break
+    case 'image':
+    case 'image_url':
+      if (content.image?.url) {
+        handlers.onImage(content.image.url)
+      }
+      break
+    case 'tool_calls':
+      if (content.tool_calls && Array.isArray(content.tool_calls)) {
+        handlers.onToolCalls(content.tool_calls)
+      }
+      break
+    case 'tool_result':
+      if (content.tool_result?.output_text?.text) {
+        handlers.onText(content.tool_result.output_text.text)
+      }
+      break
+    default:
+      // Fallback for legacy fields without explicit type
+      if (content.text?.value || content.text?.text) {
+        handlers.onText(content.text.value || content.text.text || '')
+      }
+      if (content.text_result) {
+        handlers.onText(content.text_result)
+      }
+      if (content.output_text?.text) {
+        handlers.onText(content.output_text.text)
+      }
+      if (content.reasoning_content) {
+        handlers.onReasoning(content.reasoning_content)
+      }
+      if (content.image?.url) {
+        handlers.onImage(content.image.url)
+      }
+      if (content.tool_calls && Array.isArray(content.tool_calls)) {
+        handlers.onToolCalls(content.tool_calls)
+      }
+      break
+  }
+}

extensions-web/src/jan-provider-web/api.ts

@@ -4,10 +4,11 @@
  */
 
 import { getSharedAuthService, JanAuthService } from '../shared'
-import { JanModel, janProviderStore } from './store'
 import { ApiError } from '../shared/types/errors'
+import { JAN_API_ROUTES } from './const'
+import { JanModel, janProviderStore } from './store'
 
-// JAN_API_BASE is defined in vite.config.ts
+// MENLO_PLATFORM_BASE_URL is defined in vite.config.ts
 
 // Constants
 const TEMPORARY_CHAT_ID = 'temporary-chat'
@@ -19,12 +20,7 @@ const TEMPORARY_CHAT_ID = 'temporary-chat'
  */
 function getChatCompletionConfig(request: JanChatCompletionRequest, stream: boolean = false) {
   const isTemporaryChat = request.conversation_id === TEMPORARY_CHAT_ID
-
-  // For temporary chats, use the stateless /chat/completions endpoint
-  // For regular conversations, use the stateful /conv/chat/completions endpoint
-  const endpoint = isTemporaryChat
-    ? `${JAN_API_BASE}/chat/completions`
-    : `${JAN_API_BASE}/conv/chat/completions`
+  const endpoint = `${MENLO_PLATFORM_BASE_URL}${JAN_API_ROUTES.CHAT_COMPLETIONS}`
 
   const payload = {
     ...request,
@@ -44,9 +40,30 @@ function getChatCompletionConfig(request: JanChatCompletionRequest, stream: bool
   return { endpoint, payload, isTemporaryChat }
 }
 
-export interface JanModelsResponse {
+interface JanModelSummary {
+  id: string
   object: string
-  data: JanModel[]
+  owned_by: string
+  created?: number
+}
+
+interface JanModelsResponse {
+  object: string
+  data: JanModelSummary[]
+}
+
+interface JanModelCatalogResponse {
+  id: string
+  supported_parameters?: {
+    names?: string[]
+    default?: Record<string, unknown>
+  }
+  extras?: {
+    supported_parameters?: string[]
+    default_parameters?: Record<string, unknown>
+    [key: string]: unknown
+  }
+  [key: string]: unknown
 }
 
 export interface JanChatMessage {
@@ -112,6 +129,8 @@ export interface JanChatCompletionChunk {
 export class JanApiClient {
   private static instance: JanApiClient
   private authService: JanAuthService
+  private modelsCache: JanModel[] | null = null
+  private modelsFetchPromise: Promise<JanModel[]> | null = null
 
   private constructor() {
     this.authService = getSharedAuthService()
@@ -124,25 +143,64 @@ export class JanApiClient {
     return JanApiClient.instance
   }
 
-  async getModels(): Promise<JanModel[]> {
+  async getModels(options?: { forceRefresh?: boolean }): Promise<JanModel[]> {
     try {
+      const forceRefresh = options?.forceRefresh ?? false
+
+      if (forceRefresh) {
+        this.modelsCache = null
+      } else if (this.modelsCache) {
+        return this.modelsCache
+      }
+
+      if (this.modelsFetchPromise) {
+        return this.modelsFetchPromise
+      }
+
       janProviderStore.setLoadingModels(true)
       janProviderStore.clearError()
 
-      const response = await this.authService.makeAuthenticatedRequest<JanModelsResponse>(
-        `${JAN_API_BASE}/conv/models`
-      )
+      this.modelsFetchPromise = (async () => {
+        const response = await this.authService.makeAuthenticatedRequest<JanModelsResponse>(
+          `${MENLO_PLATFORM_BASE_URL}${JAN_API_ROUTES.MODELS}`
+        )
 
-      const models = response.data || []
-      janProviderStore.setModels(models)
-      
-      return models
+        const summaries = response.data || []
+
+        const models: JanModel[] = await Promise.all(
+          summaries.map(async (summary) => {
+            const supportedParameters = await this.fetchSupportedParameters(summary.id)
+            const capabilities = this.deriveCapabilitiesFromParameters(supportedParameters)
+
+            return {
+              id: summary.id,
+              object: summary.object,
+              owned_by: summary.owned_by,
+              created: summary.created,
+              capabilities,
+              supportedParameters,
+            }
+          })
+        )
+
+        this.modelsCache = models
+        janProviderStore.setModels(models)
+
+        return models
+      })()
+
+      return await this.modelsFetchPromise
     } catch (error) {
+      this.modelsCache = null
+      this.modelsFetchPromise = null
+
       const errorMessage = error instanceof ApiError ? error.message :
                           error instanceof Error ? error.message : 'Failed to fetch models'
       janProviderStore.setError(errorMessage)
       janProviderStore.setLoadingModels(false)
       throw error
+    } finally {
+      this.modelsFetchPromise = null
     }
   }
 
@@ -254,7 +312,7 @@ export class JanApiClient {
   async initialize(): Promise<void> {
     try {
       janProviderStore.setAuthenticated(true)
-      // Fetch initial models
+      // Fetch initial models (cached for subsequent calls)
       await this.getModels()
       console.log('Jan API client initialized successfully')
     } catch (error) {
@@ -266,6 +324,52 @@ export class JanApiClient {
       janProviderStore.setInitializing(false)
     }
   }
+
+  private async fetchSupportedParameters(modelId: string): Promise<string[]> {
+    try {
+      const endpoint = `${MENLO_PLATFORM_BASE_URL}${JAN_API_ROUTES.MODEL_CATALOGS}/${this.encodeModelIdForCatalog(modelId)}`
+      const catalog = await this.authService.makeAuthenticatedRequest<JanModelCatalogResponse>(endpoint)
+      return this.extractSupportedParameters(catalog)
+    } catch (error) {
+      console.warn(`Failed to fetch catalog metadata for model "${modelId}":`, error)
+      return []
+    }
+  }
+
+  private encodeModelIdForCatalog(modelId: string): string {
+    return modelId
+      .split('/')
+      .map((segment) => encodeURIComponent(segment))
+      .join('/')
+  }
+
+  private extractSupportedParameters(catalog: JanModelCatalogResponse | null | undefined): string[] {
+    if (!catalog) {
+      return []
+    }
+
+    const primaryNames = catalog.supported_parameters?.names
+    if (Array.isArray(primaryNames) && primaryNames.length > 0) {
+      return [...new Set(primaryNames)]
+    }
+
+    const extraNames = catalog.extras?.supported_parameters
+    if (Array.isArray(extraNames) && extraNames.length > 0) {
+      return [...new Set(extraNames)]
+    }
+
+    return []
+  }
+
+  private deriveCapabilitiesFromParameters(parameters: string[]): string[] {
+    const capabilities = new Set<string>()
+
+    if (parameters.includes('tools')) {
+      capabilities.add('tools')
+    }
+
+    return Array.from(capabilities)
+  }
 }
 
 export const janApiClient = JanApiClient.getInstance()

extensions-web/src/jan-provider-web/const.ts

@@ -0,0 +1,7 @@
+export const JAN_API_ROUTES = {
+  MODELS: '/models',
+  CHAT_COMPLETIONS: '/chat/completions',
+  MODEL_CATALOGS: '/models/catalogs',
+} as const
+
+export const MODEL_PROVIDER_STORAGE_KEY = 'model-provider'

extensions-web/src/jan-provider-web/helpers.ts

@@ -0,0 +1,122 @@
+import type { JanModel } from './store'
+import { MODEL_PROVIDER_STORAGE_KEY } from './const'
+
+type StoredModel = {
+  id?: string
+  capabilities?: unknown
+  [key: string]: unknown
+}
+
+type StoredProvider = {
+  provider?: string
+  models?: StoredModel[]
+  [key: string]: unknown
+}
+
+type StoredState = {
+  state?: {
+    providers?: StoredProvider[]
+    [key: string]: unknown
+  }
+  version?: number
+  [key: string]: unknown
+}
+
+const normalizeCapabilities = (capabilities: unknown): string[] => {
+  if (!Array.isArray(capabilities)) {
+    return []
+  }
+
+  return [...new Set(capabilities.filter((item): item is string => typeof item === 'string'))].sort(
+    (a, b) => a.localeCompare(b)
+  )
+}
+
+/**
+ * Synchronize Jan models stored in localStorage with the latest server state.
+ * Returns true if the stored data was modified (including being cleared).
+ */
+export function syncJanModelsLocalStorage(
+  remoteModels: JanModel[],
+  storageKey: string = MODEL_PROVIDER_STORAGE_KEY
+): boolean {
+  const rawStorage = localStorage.getItem(storageKey)
+  if (!rawStorage) {
+    return false
+  }
+
+  let storedState: StoredState
+  try {
+    storedState = JSON.parse(rawStorage) as StoredState
+  } catch (error) {
+    console.warn('Failed to parse Jan model storage; clearing entry.', error)
+    localStorage.removeItem(storageKey)
+    return true
+  }
+
+  const providers = storedState?.state?.providers
+  if (!Array.isArray(providers)) {
+    return false
+  }
+
+  const remoteModelMap = new Map(remoteModels.map((model) => [model.id, model]))
+  let storageUpdated = false
+
+  for (const provider of providers) {
+    if (provider.provider !== 'jan' || !Array.isArray(provider.models)) {
+      continue
+    }
+
+    const updatedModels: StoredModel[] = []
+
+    for (const model of provider.models) {
+      const modelId = typeof model.id === 'string' ? model.id : null
+      if (!modelId) {
+        storageUpdated = true
+        continue
+      }
+
+      const remoteModel = remoteModelMap.get(modelId)
+      if (!remoteModel) {
+        console.log(`Removing unknown Jan model from localStorage: ${modelId}`)
+        storageUpdated = true
+        continue
+      }
+
+      const storedCapabilities = normalizeCapabilities(model.capabilities)
+      const remoteCapabilities = normalizeCapabilities(remoteModel.capabilities)
+
+      const capabilitiesMatch =
+        storedCapabilities.length === remoteCapabilities.length &&
+        storedCapabilities.every((cap, index) => cap === remoteCapabilities[index])
+
+      if (!capabilitiesMatch) {
+        console.log(
+          `Updating capabilities for Jan model ${modelId}:`,
+          storedCapabilities,
+          '=>',
+          remoteCapabilities
+        )
+        updatedModels.push({
+          ...model,
+          capabilities: remoteModel.capabilities,
+        })
+        storageUpdated = true
+      } else {
+        updatedModels.push(model)
+      }
+    }
+
+    if (updatedModels.length !== provider.models.length) {
+      storageUpdated = true
+    }
+
+    provider.models = updatedModels
+  }
+
+  if (storageUpdated) {
+    localStorage.setItem(storageKey, JSON.stringify(storedState))
+  }
+
+  return storageUpdated
+}

extensions-web/src/jan-provider-web/provider.ts

@@ -14,12 +14,10 @@ import {
   ImportOptions,
 } from '@janhq/core' // cspell: disable-line
 import { janApiClient, JanChatMessage } from './api'
+import { syncJanModelsLocalStorage } from './helpers'
 import { janProviderStore } from './store'
 import { ApiError } from '../shared/types/errors'
 
-// Jan models support tools via MCP
-const JAN_MODEL_CAPABILITIES = ['tools'] as const
-
 export default class JanProviderWeb extends AIEngine {
   readonly provider = 'jan'
   private activeSessions: Map<string, SessionInfo> = new Map()
@@ -28,11 +26,11 @@ export default class JanProviderWeb extends AIEngine {
     console.log('Loading Jan Provider Extension...')
 
     try {
-      // Check and clear invalid Jan models (capabilities mismatch)
-      this.validateJanModelsLocalStorage()
-
-      // Initialize authentication and fetch models
+      // Initialize authentication
       await janApiClient.initialize()
+      // Check and sync stored Jan models against latest catalog data
+      await this.validateJanModelsLocalStorage()
+
       console.log('Jan Provider Extension loaded successfully')
     } catch (error) {
       console.error('Failed to load Jan Provider Extension:', error)
@@ -43,46 +41,17 @@ export default class JanProviderWeb extends AIEngine {
   }
 
   // Verify Jan models capabilities in localStorage
-  private validateJanModelsLocalStorage() {
+  private async validateJanModelsLocalStorage(): Promise<void> {
     try {
-      console.log("Validating Jan models in localStorage...")
-      const storageKey = 'model-provider'
-      const data = localStorage.getItem(storageKey)
-      if (!data) return
+      console.log('Validating Jan models in localStorage...')
 
-      const parsed = JSON.parse(data)
-      if (!parsed?.state?.providers) return
+      const remoteModels = await janApiClient.getModels()
+      const storageUpdated = syncJanModelsLocalStorage(remoteModels)
 
-      // Check if any Jan model has incorrect capabilities
-      let hasInvalidModel = false
-
-      for (const provider of parsed.state.providers) {
-        if (provider.provider === 'jan' && provider.models) {
-          for (const model of provider.models) {
-            console.log(`Checking Jan model: ${model.id}`, model.capabilities)
-            if (JSON.stringify(model.capabilities) !== JSON.stringify(JAN_MODEL_CAPABILITIES)) {
-              hasInvalidModel = true
-              console.log(`Found invalid Jan model: ${model.id}, clearing localStorage`)
-              break
-            }
-          }
-        }
-        if (hasInvalidModel) break
-      }
-
-      // If any invalid model found, just clear the storage
-      if (hasInvalidModel) {
-        // Force clear the storage
-        localStorage.removeItem(storageKey)
-        // Verify it's actually removed
-        const afterRemoval = localStorage.getItem(storageKey)
-        // If still present, try setting to empty state
-        if (afterRemoval) {
-          // Try alternative clearing method
-          localStorage.setItem(storageKey, JSON.stringify({ state: { providers: [] }, version: parsed.version || 3 }))
-        }
-        console.log('Cleared model-provider from localStorage due to invalid Jan capabilities')
-        // Force a page reload to ensure clean state
+      if (storageUpdated) {
+        console.log(
+          'Synchronized Jan models in localStorage with server capabilities; reloading...'
+        )
         window.location.reload()
       }
     } catch (error) {
@@ -119,7 +88,7 @@ export default class JanProviderWeb extends AIEngine {
               path: undefined, // Remote model, no local path
               owned_by: model.owned_by,
               object: model.object,
-              capabilities: [...JAN_MODEL_CAPABILITIES],
+              capabilities: [...model.capabilities],
             }
           : undefined
       )
@@ -140,7 +109,7 @@ export default class JanProviderWeb extends AIEngine {
         path: undefined, // Remote model, no local path
         owned_by: model.owned_by,
         object: model.object,
-        capabilities: [...JAN_MODEL_CAPABILITIES],
+        capabilities: [...model.capabilities],
       }))
     } catch (error) {
       console.error('Failed to list Jan models:', error)
@@ -159,6 +128,7 @@ export default class JanProviderWeb extends AIEngine {
         port: 443, // HTTPS port
         model_id: modelId,
         model_path: `remote:${modelId}`, // Indicate this is a remote model
+        is_embedding: false, // assume false here, TODO: might need further implementation
         api_key: '', // API key handled by auth service
       }
 
@@ -193,8 +163,12 @@ export default class JanProviderWeb extends AIEngine {
       console.error(`Failed to unload Jan session ${sessionId}:`, error)
       return {
         success: false,
-        error: error instanceof ApiError ? error.message :
-               error instanceof Error ? error.message : 'Unknown error',
+        error:
+          error instanceof ApiError
+            ? error.message
+            : error instanceof Error
+              ? error.message
+              : 'Unknown error',
       }
     }
   }

extensions-web/src/jan-provider-web/store.ts

@@ -9,6 +9,9 @@ export interface JanModel {
   id: string
   object: string
   owned_by: string
+  created?: number
+  capabilities: string[]
+  supportedParameters?: string[]
 }
 
 export interface JanProviderState {

extensions-web/src/mcp-web/index.ts

@@ -12,8 +12,8 @@ import { JanMCPOAuthProvider } from './oauth-provider'
 import { WebSearchButton } from './components'
 import type { ComponentType } from 'react'
 
-// JAN_API_BASE is defined in vite.config.ts (defaults to 'https://api-dev.jan.ai/jan/v1')
-declare const JAN_API_BASE: string
+// MENLO_PLATFORM_BASE_URL is defined in vite.config.ts (defaults to 'https://api-dev.menlo.ai/jan/v1')
+declare const MENLO_PLATFORM_BASE_URL: string
 
 export default class MCPExtensionWeb extends MCPExtension {
   private mcpEndpoint = '/mcp'
@@ -77,7 +77,7 @@ export default class MCPExtensionWeb extends MCPExtension {
 
       // Create transport with OAuth provider (handles token refresh automatically)
       const transport = new StreamableHTTPClientTransport(
-        new URL(`${JAN_API_BASE}${this.mcpEndpoint}`),
+        new URL(`${MENLO_PLATFORM_BASE_URL}${this.mcpEndpoint}`),
         {
           authProvider: this.oauthProvider
           // No sessionId needed - server will generate one automatically

extensions-web/src/shared/auth/api.ts

@@ -6,13 +6,13 @@
 import { AuthTokens } from './types'
 import { AUTH_ENDPOINTS } from './const'
 
-declare const JAN_API_BASE: string
+declare const MENLO_PLATFORM_BASE_URL: string
 
 /**
  * Logout user on server
  */
 export async function logoutUser(): Promise<void> {
-  const response = await fetch(`${JAN_API_BASE}${AUTH_ENDPOINTS.LOGOUT}`, {
+  const response = await fetch(`${MENLO_PLATFORM_BASE_URL}${AUTH_ENDPOINTS.LOGOUT}`, {
     method: 'GET',
     credentials: 'include',
     headers: {
@@ -29,7 +29,7 @@ export async function logoutUser(): Promise<void> {
  * Guest login
  */
 export async function guestLogin(): Promise<AuthTokens> {
-  const response = await fetch(`${JAN_API_BASE}${AUTH_ENDPOINTS.GUEST_LOGIN}`, {
+  const response = await fetch(`${MENLO_PLATFORM_BASE_URL}${AUTH_ENDPOINTS.GUEST_LOGIN}`, {
     method: 'POST',
     credentials: 'include',
     headers: {
@@ -51,7 +51,7 @@ export async function guestLogin(): Promise<AuthTokens> {
  */
 export async function refreshToken(): Promise<AuthTokens> {
   const response = await fetch(
-    `${JAN_API_BASE}${AUTH_ENDPOINTS.REFRESH_TOKEN}`,
+    `${MENLO_PLATFORM_BASE_URL}${AUTH_ENDPOINTS.REFRESH_TOKEN}`,
     {
       method: 'GET',
       credentials: 'include',

extensions-web/src/shared/auth/providers/api.ts

@@ -5,10 +5,10 @@
 
 import { AuthTokens, LoginUrlResponse } from './types'
 
-declare const JAN_API_BASE: string
+declare const MENLO_PLATFORM_BASE_URL: string
 
 export async function getLoginUrl(endpoint: string): Promise<LoginUrlResponse> {
-  const response: Response = await fetch(`${JAN_API_BASE}${endpoint}`, {
+  const response: Response = await fetch(`${MENLO_PLATFORM_BASE_URL}${endpoint}`, {
     method: 'GET',
     credentials: 'include',
     headers: {
@@ -30,7 +30,7 @@ export async function handleOAuthCallback(
   code: string,
   state?: string
 ): Promise<AuthTokens> {
-  const response: Response = await fetch(`${JAN_API_BASE}${endpoint}`, {
+  const response: Response = await fetch(`${MENLO_PLATFORM_BASE_URL}${endpoint}`, {
     method: 'POST',
     headers: {
       'Content-Type': 'application/json',

extensions-web/src/shared/auth/service.ts

@@ -3,9 +3,9 @@
  * Handles authentication flows for any OAuth provider
  */
 
-declare const JAN_API_BASE: string
+declare const MENLO_PLATFORM_BASE_URL: string
 
-import { User, AuthState, AuthBroadcastMessage } from './types'
+import { User, AuthState, AuthBroadcastMessage, AuthTokens } from './types'
 import {
   AUTH_STORAGE_KEYS,
   AUTH_ENDPOINTS,
@@ -115,7 +115,7 @@ export class JanAuthService {
 
       // Store tokens and set authenticated state
       this.accessToken = tokens.access_token
-      this.tokenExpiryTime = Date.now() + tokens.expires_in * 1000
+      this.tokenExpiryTime = this.computeTokenExpiry(tokens)
       this.setAuthProvider(providerId)
 
       this.authBroadcast.broadcastLogin()
@@ -158,7 +158,7 @@ export class JanAuthService {
       const tokens = await refreshToken()
 
       this.accessToken = tokens.access_token
-      this.tokenExpiryTime = Date.now() + tokens.expires_in * 1000
+      this.tokenExpiryTime = this.computeTokenExpiry(tokens)
     } catch (error) {
       console.error('Failed to refresh access token:', error)
       if (error instanceof ApiError && error.isStatus(401)) {
@@ -343,6 +343,23 @@ export class JanAuthService {
     localStorage.removeItem(AUTH_STORAGE_KEYS.AUTH_PROVIDER)
   }
 
+  private computeTokenExpiry(tokens: AuthTokens): number {
+    if (tokens.expires_at) {
+      const expiresAt = new Date(tokens.expires_at).getTime()
+      if (!Number.isNaN(expiresAt)) {
+        return expiresAt
+      }
+      console.warn('Invalid expires_at format in auth tokens:', tokens.expires_at)
+    }
+
+    if (typeof tokens.expires_in === 'number') {
+      return Date.now() + tokens.expires_in * 1000
+    }
+
+    console.warn('Auth tokens missing expiry information; defaulting to immediate expiry')
+    return Date.now()
+  }
+
   /**
    * Ensure guest access is available
    */
@@ -352,7 +369,7 @@ export class JanAuthService {
       if (!this.accessToken || Date.now() > this.tokenExpiryTime) {
         const tokens = await guestLogin()
         this.accessToken = tokens.access_token
-        this.tokenExpiryTime = Date.now() + tokens.expires_in * 1000
+        this.tokenExpiryTime = this.computeTokenExpiry(tokens)
       }
     } catch (error) {
       console.error('Failed to ensure guest access:', error)
@@ -387,7 +404,6 @@ export class JanAuthService {
         case AUTH_EVENTS.LOGOUT:
           // Another tab logged out, clear our state
           this.clearAuthState()
-          this.ensureGuestAccess().catch(console.error)
           break
       }
     })
@@ -413,7 +429,7 @@ export class JanAuthService {
   private async fetchUserProfile(): Promise<User | null> {
     try {
       return await this.makeAuthenticatedRequest<User>(
-        `${JAN_API_BASE}${AUTH_ENDPOINTS.ME}`
+        `${MENLO_PLATFORM_BASE_URL}${AUTH_ENDPOINTS.ME}`
       )
     } catch (error) {
       console.error('Failed to fetch user profile:', error)

extensions-web/src/shared/auth/types.ts

@@ -16,7 +16,8 @@ export type AuthType = ProviderType | 'guest'
 
 export interface AuthTokens {
   access_token: string
-  expires_in: number
+  expires_in?: number
+  expires_at?: string
   object: string
 }
 

extensions-web/src/types/global.d.ts

@@ -1,5 +1,5 @@
 export {}
 
 declare global {
-  declare const JAN_API_BASE: string
+  declare const MENLO_PLATFORM_BASE_URL: string
 }

extensions-web/vite.config.ts

@@ -14,6 +14,6 @@ export default defineConfig({
     emptyOutDir: false // Don't clean the output directory
   },
   define: {
-    JAN_API_BASE: JSON.stringify(process.env.JAN_API_BASE || 'https://api-dev.jan.ai/v1'),
+    MENLO_PLATFORM_BASE_URL: JSON.stringify(process.env.MENLO_PLATFORM_BASE_URL || 'https://api-dev.menlo.ai/v1'),
   }
 })

extensions/assistant-extension/README.md

@@ -70,6 +70,6 @@ There are a few things to keep in mind when writing your extension code:
   ```
 
   For more information about the Jan Extension Core module, see the
-  [documentation](https://github.com/menloresearch/jan/blob/main/core/README.md).
+  [documentation](https://github.com/janhq/jan/blob/main/core/README.md).
 
 So, what are you waiting for? Go ahead and start customizing your extension!

extensions/llamacpp-extension/src/backend.ts

@@ -56,7 +56,7 @@ async function fetchRemoteSupportedBackends(
   supportedBackends: string[]
 ): Promise<{ version: string; backend: string }[]> {
   // Pull the latest releases from the repo
-  const { releases } = await _fetchGithubReleases('menloresearch', 'llama.cpp')
+  const { releases } = await _fetchGithubReleases('janhq', 'llama.cpp')
   releases.sort((a, b) => b.tag_name.localeCompare(a.tag_name))
   releases.splice(10) // keep only the latest 10 releases
 
@@ -98,7 +98,7 @@ export async function listSupportedBackends(): Promise<
   const sysType = `${os_type}-${arch}`
   let supportedBackends = []
 
-  // NOTE: menloresearch's tags for llama.cpp builds are a bit different
+  // NOTE: janhq's tags for llama.cpp builds are a bit different
   // TODO: fetch versions from the server?
   // TODO: select CUDA version based on driver version
   if (sysType == 'windows-x86_64') {
@@ -247,7 +247,7 @@ export async function downloadBackend(
   // Build URLs per source
   const backendUrl =
     source === 'github'
-      ? `https://github.com/menloresearch/llama.cpp/releases/download/${version}/llama-${version}-bin-${backend}.tar.gz`
+      ? `https://github.com/janhq/llama.cpp/releases/download/${version}/llama-${version}-bin-${backend}.tar.gz`
       : `https://catalog.jan.ai/llama.cpp/releases/${version}/llama-${version}-bin-${backend}.tar.gz`
 
   const downloadItems = [
@@ -263,7 +263,7 @@ export async function downloadBackend(
     downloadItems.push({
       url:
         source === 'github'
-          ? `https://github.com/menloresearch/llama.cpp/releases/download/${version}/cudart-llama-bin-${platformName}-cu11.7-x64.tar.gz`
+          ? `https://github.com/janhq/llama.cpp/releases/download/${version}/cudart-llama-bin-${platformName}-cu11.7-x64.tar.gz`
           : `https://catalog.jan.ai/llama.cpp/releases/${version}/cudart-llama-bin-${platformName}-cu11.7-x64.tar.gz`,
       save_path: await joinPath([libDir, 'cuda11.tar.gz']),
       proxy: proxyConfig,
@@ -272,7 +272,7 @@ export async function downloadBackend(
     downloadItems.push({
       url:
         source === 'github'
-          ? `https://github.com/menloresearch/llama.cpp/releases/download/${version}/cudart-llama-bin-${platformName}-cu12.0-x64.tar.gz`
+          ? `https://github.com/janhq/llama.cpp/releases/download/${version}/cudart-llama-bin-${platformName}-cu12.0-x64.tar.gz`
           : `https://catalog.jan.ai/llama.cpp/releases/${version}/cudart-llama-bin-${platformName}-cu12.0-x64.tar.gz`,
       save_path: await joinPath([libDir, 'cuda12.tar.gz']),
       proxy: proxyConfig,

extensions/llamacpp-extension/src/index.ts

@@ -333,14 +333,12 @@ export default class llamacpp_extension extends AIEngine {
           )
           // Clear the invalid stored preference
           this.clearStoredBackendType()
-          bestAvailableBackendString = await this.determineBestBackend(
-            version_backends
-          )
+          bestAvailableBackendString =
+            await this.determineBestBackend(version_backends)
         }
       } else {
-        bestAvailableBackendString = await this.determineBestBackend(
-          version_backends
-        )
+        bestAvailableBackendString =
+          await this.determineBestBackend(version_backends)
       }
 
       let settings = structuredClone(SETTINGS)
@@ -1530,6 +1528,7 @@ export default class llamacpp_extension extends AIEngine {
 
     if (
       this.autoUnload &&
+      !isEmbedding &&
       (loadedModels.length > 0 || otherLoadingPromises.length > 0)
     ) {
       // Wait for OTHER loading models to finish, then unload everything
@@ -1537,10 +1536,33 @@ export default class llamacpp_extension extends AIEngine {
         await Promise.all(otherLoadingPromises)
       }
 
-      // Now unload all loaded models
+      // Now unload all loaded Text models excluding embedding models
       const allLoadedModels = await this.getLoadedModels()
       if (allLoadedModels.length > 0) {
-        await Promise.all(allLoadedModels.map((model) => this.unload(model)))
+        const sessionInfos: (SessionInfo | null)[] = await Promise.all(
+          allLoadedModels.map(async (modelId) => {
+            try {
+              return await this.findSessionByModel(modelId)
+            } catch (e) {
+              logger.warn(`Unable to find session for model "${modelId}": ${e}`)
+              return null // treat as “not‑eligible for unload”
+            }
+          })
+        )
+
+        logger.info(JSON.stringify(sessionInfos))
+
+        const nonEmbeddingModels: string[] = sessionInfos
+          .filter(
+            (s): s is SessionInfo => s !== null && s.is_embedding === false
+          )
+          .map((s) => s.model_id)
+
+        if (nonEmbeddingModels.length > 0) {
+          await Promise.all(
+            nonEmbeddingModels.map((modelId) => this.unload(modelId))
+          )
+        }
       }
     }
     const args: string[] = []
@@ -1638,7 +1660,7 @@ export default class llamacpp_extension extends AIEngine {
     if (cfg.no_kv_offload) args.push('--no-kv-offload')
     if (isEmbedding) {
       args.push('--embedding')
-      args.push('--pooling mean')
+      args.push('--pooling', 'mean')
     } else {
       if (cfg.ctx_size > 0) args.push('--ctx-size', String(cfg.ctx_size))
       if (cfg.n_predict > 0) args.push('--n-predict', String(cfg.n_predict))
@@ -1677,6 +1699,7 @@ export default class llamacpp_extension extends AIEngine {
           libraryPath,
           args,
           envs,
+          isEmbedding,
         }
       )
       return sInfo
@@ -2012,6 +2035,69 @@ export default class llamacpp_extension extends AIEngine {
         libraryPath,
         envs,
       })
+      // On Linux with AMD GPUs, llama.cpp via Vulkan may report UMA (shared) memory as device-local.
+      // For clearer UX, override with dedicated VRAM from the hardware plugin when available.
+      try {
+        const sysInfo = await getSystemInfo()
+        if (sysInfo?.os_type === 'linux' && Array.isArray(sysInfo.gpus)) {
+          const usage = await getSystemUsage()
+          if (usage && Array.isArray(usage.gpus)) {
+            const uuidToUsage: Record<string, { total_memory: number; used_memory: number }> = {}
+            for (const u of usage.gpus as any[]) {
+              if (u && typeof u.uuid === 'string') {
+                uuidToUsage[u.uuid] = u
+              }
+            }
+
+            const indexToAmdUuid = new Map<number, string>()
+            for (const gpu of sysInfo.gpus as any[]) {
+              const vendorStr =
+                typeof gpu?.vendor === 'string'
+                  ? gpu.vendor
+                  : typeof gpu?.vendor === 'object' && gpu.vendor !== null
+                    ? String(gpu.vendor)
+                    : ''
+              if (
+                vendorStr.toUpperCase().includes('AMD') &&
+                gpu?.vulkan_info &&
+                typeof gpu.vulkan_info.index === 'number' &&
+                typeof gpu.uuid === 'string'
+              ) {
+                indexToAmdUuid.set(gpu.vulkan_info.index, gpu.uuid)
+              }
+            }
+
+            if (indexToAmdUuid.size > 0) {
+              const adjusted = dList.map((dev) => {
+                if (dev.id?.startsWith('Vulkan')) {
+                  const match = /^Vulkan(\d+)/.exec(dev.id)
+                  if (match) {
+                    const vIdx = Number(match[1])
+                    const uuid = indexToAmdUuid.get(vIdx)
+                    if (uuid) {
+                      const u = uuidToUsage[uuid]
+                      if (
+                        u &&
+                        typeof u.total_memory === 'number' &&
+                        typeof u.used_memory === 'number'
+                      ) {
+                        const total = Math.max(0, Math.floor(u.total_memory))
+                        const free = Math.max(0, Math.floor(u.total_memory - u.used_memory))
+                        return { ...dev, mem: total, free }
+                      }
+                    }
+                  }
+                }
+                return dev
+              })
+              return adjusted
+            }
+          }
+        }
+      } catch (e) {
+        logger.warn('Device memory override (AMD/Linux) failed:', e)
+      }
+
       return dList
     } catch (error) {
       logger.error('Failed to query devices:\n', error)
@@ -2020,6 +2106,7 @@ export default class llamacpp_extension extends AIEngine {
   }
 
   async embed(text: string[]): Promise<EmbeddingResponse> {
+    // Ensure the sentence-transformer model is present
     let sInfo = await this.findSessionByModel('sentence-transformer-mini')
     if (!sInfo) {
       const downloadedModelList = await this.list()
@@ -2033,30 +2120,45 @@ export default class llamacpp_extension extends AIEngine {
             'https://huggingface.co/second-state/All-MiniLM-L6-v2-Embedding-GGUF/resolve/main/all-MiniLM-L6-v2-ggml-model-f16.gguf?download=true',
         })
       }
-      sInfo = await this.load('sentence-transformer-mini')
+      // Load specifically in embedding mode
+      sInfo = await this.load('sentence-transformer-mini', undefined, true)
     }
-    const baseUrl = `http://localhost:${sInfo.port}/v1/embeddings`
-    const headers = {
-      'Content-Type': 'application/json',
-      'Authorization': `Bearer ${sInfo.api_key}`,
+
+    const attemptRequest = async (session: SessionInfo) => {
+      const baseUrl = `http://localhost:${session.port}/v1/embeddings`
+      const headers = {
+        'Content-Type': 'application/json',
+        'Authorization': `Bearer ${session.api_key}`,
+      }
+      const body = JSON.stringify({
+        input: text,
+        model: session.model_id,
+        encoding_format: 'float',
+      })
+      const response = await fetch(baseUrl, {
+        method: 'POST',
+        headers,
+        body,
+      })
+      return response
+    }
+
+    // First try with the existing session (may have been started without --embedding previously)
+    let response = await attemptRequest(sInfo)
+
+    // If embeddings endpoint is not available (501), reload with embedding mode and retry once
+    if (response.status === 501) {
+      try {
+        await this.unload('sentence-transformer-mini')
+      } catch {}
+      sInfo = await this.load('sentence-transformer-mini', undefined, true)
+      response = await attemptRequest(sInfo)
     }
-    const body = JSON.stringify({
-      input: text,
-      model: sInfo.model_id,
-      encoding_format: 'float',
-    })
-    const response = await fetch(baseUrl, {
-      method: 'POST',
-      headers,
-      body,
-    })
 
     if (!response.ok) {
       const errorData = await response.json().catch(() => null)
       throw new Error(
-        `API request failed with status ${response.status}: ${JSON.stringify(
-          errorData
-        )}`
+        `API request failed with status ${response.status}: ${JSON.stringify(errorData)}`
       )
     }
     const responseData = await response.json()

extensions/rag-extension/package.json

@@ -0,0 +1,33 @@
+{
+  "name": "@janhq/rag-extension",
+  "productName": "RAG Tools",
+  "version": "0.1.0",
+  "description": "Registers RAG tools and orchestrates retrieval across parser, embeddings, and vector DB",
+  "main": "dist/index.js",
+  "module": "dist/module.js",
+  "author": "Jan <service@jan.ai>",
+  "license": "AGPL-3.0",
+  "scripts": {
+    "build": "rolldown -c rolldown.config.mjs",
+    "build:publish": "rimraf *.tgz --glob || true && yarn build && npm pack && cpx *.tgz ../../pre-install"
+  },
+  "devDependencies": {
+    "cpx": "1.5.0",
+    "rimraf": "6.0.1",
+    "rolldown": "1.0.0-beta.1",
+    "typescript": "5.9.2"
+  },
+  "dependencies": {
+    "@janhq/core": "../../core/package.tgz",
+    "@janhq/tauri-plugin-rag-api": "link:../../src-tauri/plugins/tauri-plugin-rag",
+    "@janhq/tauri-plugin-vector-db-api": "link:../../src-tauri/plugins/tauri-plugin-vector-db"
+  },
+  "files": [
+    "dist/*",
+    "package.json"
+  ],
+  "installConfig": {
+    "hoistingLimits": "workspaces"
+  },
+  "packageManager": "yarn@4.5.3"
+}

extensions/rag-extension/rolldown.config.mjs

@@ -0,0 +1,14 @@
+import { defineConfig } from 'rolldown'
+import settingJson from './settings.json' with { type: 'json' }
+
+export default defineConfig({
+  input: 'src/index.ts',
+  output: {
+    format: 'esm',
+    file: 'dist/index.js',
+  },
+  platform: 'browser',
+  define: {
+    SETTINGS: JSON.stringify(settingJson),
+  },
+})

extensions/rag-extension/settings.json

@@ -0,0 +1,58 @@
+[
+  {
+    "key": "enabled",
+    "titleKey": "settings:attachments.enable",
+    "descriptionKey": "settings:attachments.enableDesc",
+    "controllerType": "checkbox",
+    "controllerProps": { "value": true }
+  },
+  {
+    "key": "max_file_size_mb",
+    "titleKey": "settings:attachments.maxFile",
+    "descriptionKey": "settings:attachments.maxFileDesc",
+    "controllerType": "input",
+    "controllerProps": { "value": 20, "type": "number", "min": 1, "max": 200, "step": 1, "textAlign": "right" }
+  },
+  {
+    "key": "retrieval_limit",
+    "titleKey": "settings:attachments.topK",
+    "descriptionKey": "settings:attachments.topKDesc",
+    "controllerType": "input",
+    "controllerProps": { "value": 3, "type": "number", "min": 1, "max": 20, "step": 1, "textAlign": "right" }
+  },
+  {
+    "key": "retrieval_threshold",
+    "titleKey": "settings:attachments.threshold",
+    "descriptionKey": "settings:attachments.thresholdDesc",
+    "controllerType": "input",
+    "controllerProps": { "value": 0.3, "type": "number", "min": 0, "max": 1, "step": 0.01, "textAlign": "right" }
+  },
+  {
+    "key": "chunk_size_tokens",
+    "titleKey": "settings:attachments.chunkSize",
+    "descriptionKey": "settings:attachments.chunkSizeDesc",
+    "controllerType": "input",
+    "controllerProps": { "value": 512, "type": "number", "min": 64, "max": 8192, "step": 64, "textAlign": "right" }
+  },
+  {
+    "key": "overlap_tokens",
+    "titleKey": "settings:attachments.chunkOverlap",
+    "descriptionKey": "settings:attachments.chunkOverlapDesc",
+    "controllerType": "input",
+    "controllerProps": { "value": 64, "type": "number", "min": 0, "max": 1024, "step": 16, "textAlign": "right" }
+  },
+  {
+    "key": "search_mode",
+    "titleKey": "settings:attachments.searchMode",
+    "descriptionKey": "settings:attachments.searchModeDesc",
+    "controllerType": "dropdown",
+    "controllerProps": {
+      "value": "auto",
+      "options": [
+        { "name": "Auto (recommended)", "value": "auto" },
+        { "name": "ANN (sqlite-vec)", "value": "ann" },
+        { "name": "Linear", "value": "linear" }
+      ]
+    }
+  }
+]

extensions/rag-extension/src/env.d.ts

@@ -0,0 +1,5 @@
+import type { SettingComponentProps } from '@janhq/core'
+declare global {
+  const SETTINGS: SettingComponentProps[]
+}
+export {}

extensions/rag-extension/src/global.d.ts

@@ -0,0 +1,14 @@
+import type { BaseExtension, ExtensionTypeEnum } from '@janhq/core'
+
+declare global {
+  interface Window {
+    core?: {
+      extensionManager: {
+        get<T = BaseExtension>(type: ExtensionTypeEnum): T | undefined
+        getByName(name: string): BaseExtension | undefined
+      }
+    }
+  }
+}
+
+export {}

extensions/rag-extension/src/index.ts

@@ -0,0 +1,305 @@
+import { RAGExtension, MCPTool, MCPToolCallResult, ExtensionTypeEnum, VectorDBExtension, type AttachmentInput, type SettingComponentProps, AIEngine, type AttachmentFileInfo } from '@janhq/core'
+import './env.d'
+import { getRAGTools, RETRIEVE, LIST_ATTACHMENTS, GET_CHUNKS } from './tools'
+
+export default class RagExtension extends RAGExtension {
+  private config = {
+    enabled: true,
+    retrievalLimit: 3,
+    retrievalThreshold: 0.3,
+    chunkSizeTokens: 512,
+    overlapTokens: 64,
+    searchMode: 'auto' as 'auto' | 'ann' | 'linear',
+    maxFileSizeMB: 20,
+  }
+
+  async onLoad(): Promise<void> {
+    const settings = structuredClone(SETTINGS) as SettingComponentProps[]
+    await this.registerSettings(settings)
+    this.config.enabled = await this.getSetting('enabled', this.config.enabled)
+    this.config.maxFileSizeMB = await this.getSetting('max_file_size_mb', this.config.maxFileSizeMB)
+    this.config.retrievalLimit = await this.getSetting('retrieval_limit', this.config.retrievalLimit)
+    this.config.retrievalThreshold = await this.getSetting('retrieval_threshold', this.config.retrievalThreshold)
+    this.config.chunkSizeTokens = await this.getSetting('chunk_size_tokens', this.config.chunkSizeTokens)
+    this.config.overlapTokens = await this.getSetting('overlap_tokens', this.config.overlapTokens)
+    this.config.searchMode = await this.getSetting('search_mode', this.config.searchMode)
+
+    // Check ANN availability on load
+    try {
+      const vec = window.core?.extensionManager.get(ExtensionTypeEnum.VectorDB) as unknown as VectorDBExtension
+      if (vec?.getStatus) {
+        const status = await vec.getStatus()
+        console.log('[RAG] Vector DB ANN support:', status.ann_available ? '✓ AVAILABLE' : '✗ NOT AVAILABLE')
+        if (!status.ann_available) {
+          console.warn('[RAG] Warning: sqlite-vec not loaded. Collections will use slower linear search.')
+        }
+      }
+    } catch (e) {
+      console.error('[RAG] Failed to check ANN status:', e)
+    }
+  }
+
+  onUnload(): void {}
+
+  async getTools(): Promise<MCPTool[]> {
+    return getRAGTools(this.config.retrievalLimit)
+  }
+
+  async getToolNames(): Promise<string[]> {
+    // Keep this in sync with getTools() but without building full schemas
+    return [LIST_ATTACHMENTS, RETRIEVE, GET_CHUNKS]
+  }
+
+  async callTool(toolName: string, args: Record<string, unknown>): Promise<MCPToolCallResult> {
+    switch (toolName) {
+      case LIST_ATTACHMENTS:
+        return this.listAttachments(args)
+      case RETRIEVE:
+        return this.retrieve(args)
+      case GET_CHUNKS:
+        return this.getChunks(args)
+      default:
+        return {
+          error: `Unknown tool: ${toolName}`,
+          content: [{ type: 'text', text: `Unknown tool: ${toolName}` }],
+        }
+    }
+  }
+
+  private async listAttachments(args: Record<string, unknown>): Promise<MCPToolCallResult> {
+    const threadId = String(args['thread_id'] || '')
+    if (!threadId) {
+      return { error: 'Missing thread_id', content: [{ type: 'text', text: 'Missing thread_id' }] }
+    }
+    try {
+      const vec = window.core?.extensionManager.get(ExtensionTypeEnum.VectorDB) as unknown as VectorDBExtension
+      if (!vec?.listAttachments) {
+        return { error: 'Vector DB extension missing listAttachments', content: [{ type: 'text', text: 'Vector DB extension missing listAttachments' }] }
+      }
+      const files = await vec.listAttachments(threadId)
+      return {
+        error: '',
+        content: [
+          {
+            type: 'text',
+            text: JSON.stringify({ thread_id: threadId, attachments: files || [] }),
+          },
+        ],
+      }
+    } catch (e) {
+      const msg = e instanceof Error ? e.message : String(e)
+      return { error: msg, content: [{ type: 'text', text: `List attachments failed: ${msg}` }] }
+    }
+  }
+
+  private async retrieve(args: Record<string, unknown>): Promise<MCPToolCallResult> {
+    const threadId = String(args['thread_id'] || '')
+    const query = String(args['query'] || '')
+    const fileIds = args['file_ids'] as string[] | undefined
+
+    const s = this.config
+    const topK = (args['top_k'] as number) || s.retrievalLimit || 3
+    const threshold = s.retrievalThreshold ?? 0.3
+    const mode: 'auto' | 'ann' | 'linear' = s.searchMode || 'auto'
+
+    if (s.enabled === false) {
+      return {
+        error: 'Attachments feature disabled',
+        content: [
+          {
+            type: 'text',
+            text: 'Attachments are disabled in Settings. Enable them to use retrieval.',
+          },
+        ],
+      }
+    }
+    if (!threadId || !query) {
+      return {
+        error: 'Missing thread_id or query',
+        content: [{ type: 'text', text: 'Missing required parameters' }],
+      }
+    }
+
+    try {
+      // Resolve extensions
+      const vec = window.core?.extensionManager.get(ExtensionTypeEnum.VectorDB) as unknown as VectorDBExtension
+      if (!vec?.searchCollection) {
+        return {
+          error: 'RAG dependencies not available',
+          content: [
+            { type: 'text', text: 'Vector DB extension not available' },
+          ],
+        }
+      }
+
+      const queryEmb = (await this.embedTexts([query]))?.[0]
+      if (!queryEmb) {
+        return {
+          error: 'Failed to compute embeddings',
+          content: [{ type: 'text', text: 'Failed to compute embeddings' }],
+        }
+      }
+
+      const results = await vec.searchCollection(
+        threadId,
+        queryEmb,
+        topK,
+        threshold,
+        mode,
+        fileIds
+      )
+
+      const payload = {
+        thread_id: threadId,
+        query,
+        citations: results?.map((r: any) => ({
+          id: r.id,
+          text: r.text,
+          score: r.score,
+          file_id: r.file_id,
+          chunk_file_order: r.chunk_file_order
+        })) ?? [],
+        mode,
+      }
+      return { error: '', content: [{ type: 'text', text: JSON.stringify(payload) }] }
+    } catch (e) {
+      console.error('[RAG] Retrieve error:', e)
+      let msg = 'Unknown error'
+      if (e instanceof Error) {
+        msg = e.message
+      } else if (typeof e === 'string') {
+        msg = e
+      } else if (e && typeof e === 'object') {
+        msg = JSON.stringify(e)
+      }
+      return { error: msg, content: [{ type: 'text', text: `Retrieve failed: ${msg}` }] }
+    }
+  }
+
+  private async getChunks(args: Record<string, unknown>): Promise<MCPToolCallResult> {
+    const threadId = String(args['thread_id'] || '')
+    const fileId = String(args['file_id'] || '')
+    const startOrder = args['start_order'] as number | undefined
+    const endOrder = args['end_order'] as number | undefined
+
+    if (!threadId || !fileId || startOrder === undefined || endOrder === undefined) {
+      return {
+        error: 'Missing thread_id, file_id, start_order, or end_order',
+        content: [{ type: 'text', text: 'Missing required parameters' }],
+      }
+    }
+
+    try {
+      const vec = window.core?.extensionManager.get(ExtensionTypeEnum.VectorDB) as unknown as VectorDBExtension
+      if (!vec?.getChunks) {
+        return {
+          error: 'Vector DB extension not available',
+          content: [{ type: 'text', text: 'Vector DB extension not available' }],
+        }
+      }
+
+      const chunks = await vec.getChunks(threadId, fileId, startOrder, endOrder)
+
+      const payload = {
+        thread_id: threadId,
+        file_id: fileId,
+        chunks: chunks || [],
+      }
+      return { error: '', content: [{ type: 'text', text: JSON.stringify(payload) }] }
+    } catch (e) {
+      const msg = e instanceof Error ? e.message : String(e)
+      return { error: msg, content: [{ type: 'text', text: `Get chunks failed: ${msg}` }] }
+    }
+  }
+
+  // Desktop-only ingestion by file paths
+  async ingestAttachments(
+    threadId: string,
+    files: AttachmentInput[]
+  ): Promise<{ filesProcessed: number; chunksInserted: number; files: AttachmentFileInfo[] }> {
+    if (!threadId || !Array.isArray(files) || files.length === 0) {
+      return { filesProcessed: 0, chunksInserted: 0, files: [] }
+    }
+
+    // Respect feature flag: do nothing when disabled
+    if (this.config.enabled === false) {
+      return { filesProcessed: 0, chunksInserted: 0, files: [] }
+    }
+
+    const vec = window.core?.extensionManager.get(ExtensionTypeEnum.VectorDB) as unknown as VectorDBExtension
+    if (!vec?.createCollection || !vec?.insertChunks) {
+      throw new Error('Vector DB extension not available')
+    }
+
+    // Load settings
+    const s = this.config
+    const maxSize = (s?.enabled === false ? 0 : s?.maxFileSizeMB) || undefined
+    const chunkSize = s?.chunkSizeTokens as number | undefined
+    const chunkOverlap = s?.overlapTokens as number | undefined
+
+    let totalChunks = 0
+    const processedFiles: AttachmentFileInfo[] = []
+
+    for (const f of files) {
+      if (!f?.path) continue
+      if (maxSize && f.size && f.size > maxSize * 1024 * 1024) {
+        throw new Error(`File '${f.name}' exceeds size limit (${f.size} bytes > ${maxSize} MB).`)
+      }
+
+      const fileName = f.name || f.path.split(/[\\/]/).pop()
+      // Preferred/required path: let Vector DB extension handle full file ingestion
+      const canIngestFile = typeof (vec as any)?.ingestFile === 'function'
+      if (!canIngestFile) {
+        console.error('[RAG] Vector DB extension missing ingestFile; cannot ingest document')
+        continue
+      }
+      const info = await (vec as VectorDBExtension).ingestFile(
+        threadId,
+        { path: f.path, name: fileName, type: f.type, size: f.size },
+        { chunkSize: chunkSize ?? 512, chunkOverlap: chunkOverlap ?? 64 }
+      )
+      totalChunks += Number(info?.chunk_count || 0)
+      processedFiles.push(info)
+    }
+
+    // Return files we ingested with real IDs directly from ingestFile
+    return { filesProcessed: processedFiles.length, chunksInserted: totalChunks, files: processedFiles }
+  }
+
+  onSettingUpdate<T>(key: string, value: T): void {
+    switch (key) {
+      case 'enabled':
+        this.config.enabled = Boolean(value)
+        break
+      case 'max_file_size_mb':
+        this.config.maxFileSizeMB = Number(value)
+        break
+      case 'retrieval_limit':
+        this.config.retrievalLimit = Number(value)
+        break
+      case 'retrieval_threshold':
+        this.config.retrievalThreshold = Number(value)
+        break
+      case 'chunk_size_tokens':
+        this.config.chunkSizeTokens = Number(value)
+        break
+      case 'overlap_tokens':
+        this.config.overlapTokens = Number(value)
+        break
+      case 'search_mode':
+        this.config.searchMode = String(value) as 'auto' | 'ann' | 'linear'
+        break
+    }
+  }
+
+  // Locally implement embedding logic (previously in embeddings-extension)
+  private async embedTexts(texts: string[]): Promise<number[][]> {
+    const llm = window.core?.extensionManager.getByName('@janhq/llamacpp-extension') as AIEngine & { embed?: (texts: string[]) => Promise<{ data: Array<{ embedding: number[]; index: number }> }> }
+    if (!llm?.embed) throw new Error('llamacpp extension not available')
+    const res = await llm.embed(texts)
+    const data: Array<{ embedding: number[]; index: number }> = res?.data || []
+    const out: number[][] = new Array(texts.length)
+    for (const item of data) out[item.index] = item.embedding
+    return out
+  }
+}

extensions/rag-extension/src/tools.ts

@@ -0,0 +1,58 @@
+import { MCPTool, RAG_INTERNAL_SERVER } from '@janhq/core'
+
+// Tool names
+export const RETRIEVE = 'retrieve'
+export const LIST_ATTACHMENTS = 'list_attachments'
+export const GET_CHUNKS = 'get_chunks'
+
+export function getRAGTools(retrievalLimit: number): MCPTool[] {
+  const maxTopK = Math.max(1, Number(retrievalLimit ?? 3))
+
+  return [
+    {
+      name: LIST_ATTACHMENTS,
+      description:
+        'List files attached to the current thread. Thread is inferred automatically; you may optionally provide {"scope":"thread"}. Returns basic file info (name/path).',
+      inputSchema: {
+        type: 'object',
+        properties: {
+          scope: { type: 'string', enum: ['thread'], description: 'Retrieval scope; currently only thread is supported' },
+        },
+        required: ['scope'],
+      },
+      server: RAG_INTERNAL_SERVER,
+    },
+    {
+      name: RETRIEVE,
+      description:
+        'Retrieve relevant snippets from locally attached, indexed documents. Use query only; do not pass raw document content. Thread context is inferred automatically; you may optionally provide {"scope":"thread"}. Use file_ids to search within specific files only.',
+      inputSchema: {
+        type: 'object',
+        properties: {
+          query: { type: 'string', description: 'User query to search for' },
+          top_k: { type: 'number', description: 'Optional: Max citations to return. Adjust as needed.', minimum: 1, maximum: maxTopK, default: retrievalLimit ?? 3 },
+          scope: { type: 'string', enum: ['thread'], description: 'Retrieval scope; currently only thread is supported' },
+          file_ids: { type: 'array', items: { type: 'string' }, description: 'Optional: Filter search to specific file IDs from list_attachments' },
+        },
+        required: ['query', 'scope'],
+      },
+      server: RAG_INTERNAL_SERVER,
+    },
+    {
+      name: GET_CHUNKS,
+      description:
+        'Retrieve chunks from a file by their order range. For a single chunk, use start_order = end_order. Thread context is inferred automatically; you may optionally provide {"scope":"thread"}. Use sparingly; intended for advanced usage. Prefer using retrieve instead for relevance-based fetching.',
+      inputSchema: {
+        type: 'object',
+        properties: {
+          file_id: { type: 'string', description: 'File ID from list_attachments' },
+          start_order: { type: 'number', description: 'Start of chunk range (inclusive, 0-indexed)' },
+          end_order: { type: 'number', description: 'End of chunk range (inclusive, 0-indexed). For single chunk, use start_order = end_order.' },
+          scope: { type: 'string', enum: ['thread'], description: 'Retrieval scope; currently only thread is supported' },
+        },
+        required: ['file_id', 'start_order', 'end_order', 'scope'],
+      },
+      server: RAG_INTERNAL_SERVER,
+    },
+  ]
+}

extensions/vector-db-extension/package.json

@@ -0,0 +1,33 @@
+{
+  "name": "@janhq/vector-db-extension",
+  "productName": "Vector DB",
+  "version": "0.1.0",
+  "description": "Vector DB integration using sqlite-vec if available with linear fallback",
+  "main": "dist/index.js",
+  "module": "dist/module.js",
+  "author": "Jan <service@jan.ai>",
+  "license": "AGPL-3.0",
+  "scripts": {
+    "build": "rolldown -c rolldown.config.mjs",
+    "build:publish": "rimraf *.tgz --glob || true && yarn build && npm pack && cpx *.tgz ../../pre-install"
+  },
+  "devDependencies": {
+    "cpx": "1.5.0",
+    "rimraf": "6.0.1",
+    "rolldown": "1.0.0-beta.1",
+    "typescript": "5.9.2"
+  },
+  "dependencies": {
+    "@janhq/core": "../../core/package.tgz",
+    "@janhq/tauri-plugin-rag-api": "link:../../src-tauri/plugins/tauri-plugin-rag",
+    "@janhq/tauri-plugin-vector-db-api": "link:../../src-tauri/plugins/tauri-plugin-vector-db"
+  },
+  "files": [
+    "dist/*",
+    "package.json"
+  ],
+  "installConfig": {
+    "hoistingLimits": "workspaces"
+  },
+  "packageManager": "yarn@4.5.3"
+}

extensions/vector-db-extension/rolldown.config.mjs

@@ -0,0 +1,11 @@
+import { defineConfig } from 'rolldown'
+
+export default defineConfig({
+  input: 'src/index.ts',
+  output: {
+    format: 'esm',
+    file: 'dist/index.js',
+  },
+  platform: 'browser',
+  define: {},
+})