Attu v3.0.1

Changes since v3.0.0.

🐛 Bug Fixes

• update vulnerable build dependencies (91945f48)
• reap detached background services (431486f7)
• allow private service endpoints by default (0456a2f5)
• align Attu volume permissions (910b2599)

🔧 CI/CD

• use bash for desktop SBOM generation (011b0c9c)
• resolve backup metadata on Windows (3a673d90)

Downloads

• macOS: .dmg (Apple Silicon)
• Linux: .deb / .AppImage
• Windows: .exe installer
• Server: attu-server-3.0.1-linux-x64-node24.tar.gz

Server tarball

The server tarball is for running Attu without Docker or Electron on Linux x64.
Node.js 24.x is required. The tarball includes the Attu server bundle and bin/milvus-backup.

tar -xzf attu-server-3.0.1-linux-x64-node24.tar.gz
cd attu-server-3.0.1-linux-x64-node24
./bin/attu-server

By default, the server listens on 0.0.0.0:3080 and stores local data under ./data.
Override runtime settings with environment variables:

HOST=127.0.0.1 PORT=4080 ATTU_DATA_DIR=/var/lib/attu ./bin/attu-server

Docker

docker run -d --name attu -p 3000:3000 -v attu-data:/data zilliz/attu:v3.0.1

Attu 3.0

Attu 3.0 delivers a richer visual experience for Milvus 3.0, with support for lakehouse data access, online schema evolution, long-text fields, advanced search analytics, reranking, and document deduplication.

Learn more about Milvus 3.0.

Highlights

External Collections and Snapshots

Explore Parquet and Iceberg data directly from your data lake without copying it into Milvus. Attu supports creating External Collections, mapping fields, refreshing source data, tracking refresh progress, and querying external data.

Snapshot management now includes browsing, restoring, and monitoring Collection restore tasks with clearer progress and error reporting.

Text Fields and Online Schema Evolution

Attu supports Milvus 3.0 TEXT fields across Collection creation, Schema, Query, Search, BM25, and analyzer workflows.

You can also evolve Collection schemas directly in Attu:

• Add or delete fields when supported
• Edit Collection and field descriptions
• Create and inspect Function Fields
• Manage schema actions based on the connected Milvus version

Attu automatically checks Storage V3 requirements and provides configuration guidance when needed.

Search and Analytics

Query and Search now provide more powerful analytics:

• COUNT, SUM, AVG, MIN, and MAX
• Multi-field grouping and ordering
• Bucket-based vector search aggregation
• Paginated Query and Search results
• Interactive table sorting
• Improved filter expression suggestions

Function Score reranking lets you combine vector similarity with scalar signals such as freshness, popularity, or quality.

Learn more about Search Aggregation.

MinHash Document Deduplication

Create MinHash Function Fields and compatible MINHASH_LSH indexes directly in Attu. You can search using raw text to identify near-duplicate documents for deduplication, clustering, and similarity analysis.

Learn more about MinHash.

Analyzers and File Resources

The new File Resources page provides centralized management for analyzer resources such as custom dictionaries, stop-word lists, synonym lists, and tokenizer files. Registered resources can be reused across multiple Collections.

Collection and Data Management

Attu 3.0 also adds:

• Partition-level load and release
• Entity-level TTL configuration
• Database and Collection property editing
• Safer deletion and truncation workflows
• Version-aware index options
• Complete Collection export
• More reliable large-file imports with chunked uploads

Compatibility

• Supports Zilliz Cloud, Milvus 2.6.0+, and Milvus 3.x.
• External Collections, Snapshots, Text fields, MinHash, entity-level TTL, and some schema operations require Milvus 3.x.
• Text fields and External Collections require Storage V3. If it is disabled, set common.storage.useLoonFFI=true and restart Milvus.
• External Collection data remains read-only in Attu. Insert, Upsert, Delete, Import, and Truncate operations are unavailable.

For detailed documentation, visit the Milvus documentation.

Attu v3.0.0-beta.6

This release summarizes the changes since v3.0.0-beta.5.

Highlights

• Added local authentication with user isolation for the app runtime, including safer password recovery handling and an 8-character minimum password length.
• Added admin connection management, including an admin view for managing connections and exporting selected admin connections.
• Added audit logs, including an admin view across all audit logs and UI refinements for the audit log table and filters.
• Added a standalone Linux x64 server tarball for running Attu without Docker or Electron.
• Improved backup and restore support by bundling milvus-backup in Docker and standalone builds, isolating backup process environment variables, and improving object storage deep-check behavior.

Explorer and Collection Management

• Added a collection index management tab.
• Added a collection segments explorer tab, with clearer segment load and query index status.
• Added shard count support when creating collections.
• Added search from existing data row vectors.
• Improved collection tree behavior, including selected collection scrolling, loaded collection highlighting, moved route handling, and refresh after collection copy.
• Improved partition selection, sample import handling, row count refresh, entity count formatting, and vector search top-k bounds.
• Fixed stale explorer routes, collection rename warnings, active collection drop warnings, long collection names in headers, imported index params, and index label translations.

Backup, Diagnostics, and Storage

• Improved storage configuration testing and isolated backup storage deep checks.
• Added diagnostic capacity scenarios and clearer diagnostic resource estimates.
• Optimized diagnose collection scanning.
• Scoped snapshot restore jobs correctly to the source collection and improved restore job feedback.

App and Runtime

• Improved connection auth UI and added connection import/export.
• Improved settings navigation and sidebar accessibility.
• Adjusted cluster navigation support links and added the Milvus Office Hour sidebar link.
• Fixed Electron cluster overview loading and optimized Electron startup.
• Repaired partial database migrations.
• Updated Milvus SDK dependencies and milvus-backup to 0.5.15.

Packaging and Release

• Added the Linux x64 standalone server tarball artifact: attu-server-3.0.0-beta.6-linux-x64-node24.tar.gz.
• Fixed standalone tarball packaging so it no longer contains absolute symlinks.
• Fixed Docker packaging so milvus-backup is available for backup deep checks.
• Fixed parquet parser packaging by using the declared dependency.

Downloads

• macOS: .dmg / .zip (Apple Silicon)
• Linux desktop: .deb / .AppImage
• Windows: .exe installer / .zip
• Standalone server: attu-server-3.0.0-beta.6-linux-x64-node24.tar.gz

Standalone Server

The standalone server tarball is for running Attu without Docker or Electron on Linux x64.
Node.js 24.x is required. The tarball includes the Attu server bundle and bin/milvus-backup.

tar -xzf attu-server-3.0.0-beta.6-linux-x64-node24.tar.gz
cd attu-server-3.0.0-beta.6-linux-x64-node24
./bin/attu-server

By default, the server listens on 0.0.0.0:3080 and stores local data under ./data.
Override runtime settings with environment variables:

HOST=127.0.0.1 PORT=4080 ATTU_DATA_DIR=/var/lib/attu ./bin/attu-server

Docker

docker run -d --name attu -p 3000:3000 -v attu-data:/data zilliz/attu:v3.0.0-beta.6

Attu v3.0.0-beta.5

Changes since v3.0.0-beta.4.

✨ New Features

• add search order by (93f9575f)
• support external collections (9ea91938)

🐛 Bug Fixes

• persist docker data under mounted volume (49e86feb)
• refresh backup process on storage changes (f912b725)
• scope backup operations by database (#98) (ea9921d1)
• avoid bigint in data explorer (85dd276d)
• gate external collections by milvus version (8cddd2a0)
• disable copy for external collections (e50611d3)
• scope milvus clients by database (013b8c8b)
• use index metric for vector search (0068eade)
• disable external collection exports (a44cc158)
• guard stale explorer routes (533945d9)

Downloads

• macOS: .dmg (Apple Silicon)
• Linux: .deb / .AppImage
• Windows: .exe installer

Docker

docker run -d --name attu -p 3000:3000 -v attu-data:/data zilliz/attu:v3.0.0-beta.5

Attu v3.0.0-beta.4

✨ New Features

• support collection snapshots (d7ca6b46)
• support nullable vector fields (14292804)
• support custom role privileges (bc17133e)
• support TLS Milvus connections (40272f63)
• show collection row counts in explorer (0c7f1553)
• support clustering key in collection schema (d94d85bb)
• allow root user to change password (a4346ad1)
• prefill num_partitions when duplicating collection (31ab8eb4)
• surface num_partitions in CollectionSchema (16dc9c45)
• pass num_partitions through createCollectionV2 to SDK (2b0b45ae)
• add resolveNumPartitions helper + unit tests (ac8ed255)
• add num_partitions input under partition key toggle (2c04074a)
• add num_partitions i18n keys (zh) (2d80d794)
• add num_partitions i18n keys (en) (7fe2aadd)
• add num_partitions field to FieldTreeNode type (58ce2334)
• include OS and deploy mode in support tickets (c7d93ec3)
• clear search results when all vector inputs emptied (857e615d)
• zero-config preview on first collection visit (3086d0ac)
• show highlight in table view too (bc78c3c6)
• add search highlight for BM25 results (1ad4eb22)
• add collapsible search panel (9e2165fd)
• add i18n for all preview components (en + zh) (c3525517)
• wire preview mode into fullscreen search results (e1576acd)
• add preview config sheet with mapping and facet tabs (d01906a6)
• add preview detail sheet with keyboard nav (094fda5f)
• add facet sidebar with checkbox/range/toggle facets (38eb170c)
• add preview result layouts and card component (33b19a9e)
• scaffold PreviewShell and PreviewTopBar (8baeed6b)
• add preview filter builder and usePreviewFilters hook (36a4e2f7)
• add usePreviewConfig and usePreviewLayout hooks (43ad03bf)
• add scanFacetCandidates server function (698da4c6)
• add savePreviewConfig server function (8b918ae2)
• add initializePreview and sampleScalarValues (7b2464ce)
• add validateConfigAgainstSchema self-healing (a15fba01)
• add isPreviewable helper (6a7e3250)
• add inferFieldMapping with fixture-driven tests (69e0300c)
• add preview-types module (9a45217e)
• add collection_display_config table for preview mode (c850facc)
• show collection name in fullscreen results header (103dd1e1)
• add right-click context menus to database tree (adbc502a)
• persist search query and results across reloads (9659a872)
• redesign dialog layout and support collection description (dd8d8fef)
• localize date/time display per current language (38712cfb)
• remember last visited tab per nav section (a3fe492e)

🐛 Bug Fixes

• disable data export for unloaded collections (40d656dd)
• allow cloud role creation (614c5571)
• hide tls switch for zilliz cloud connections (593428f5)
• repair partial TLS migrations (3e060b93)
• preserve collection database on rename (9406434f)
• keep explorer display config server-only (82f0f75a)
• cap segment buffer at actual data size for tiny collections (c144367c)
• reset search state and cleanup display config on collection lifecycle (9f9c24b5)
• count loaded collections via showCollections type=Loaded (f977d509)
• show actual num_partitions on partition key field (ef421ef1)
• data tab stays empty after reload if prior query failed (1a7f0642)
• analyzer test returns empty + flashes stale tokens (c95071a8)
• allow choosing install directory on Windows (23ab13cd)
• search-engine-style highlight snippet + deduplication (b6a2fd43)
• highlight fallback — check title and any field, not just description (9afe21e3)
• address review findings — render-time setState, stale refs, stack overflow (185d454f)
• wire facet filter into actual Milvus search pipeline (91fa1cf3)
• show facet sidebar and config in all layouts, not just card (6dc8234e)
• prevent duplicate search on facet change (901ce42f)
• connect facet sidebar filters to Milvus search (ddb9e8fe)
• fix preview search infinite loop and stabilize callbacks (827bad90)
• connect preview search box to real Milvus search (ad787d89)
• auto-create vector request for single-vector collections (c3f5deae)
• keep rerank Select trigger label clean (f2feec72)
• normalize Milvus function type from string enum (c65f8184)
• use BM25 metric for BM25 function output fields (9bcd78d1)
• add missing zh translations for scoped access and users tab (5bfbf96d)
• translate load state and align Milvus terms with Zilliz zh docs (bcf2d69c)
• accept legacy Attu v2 JSON export on import (61f5bba3)
• auto-update restart with manual fallback for unsigned builds (a7ddceb7)

🔧 CI/CD

• use gh release edit to preserve draft artifacts and fix changelog range (65cb1937)
• improve app release changelog generation (f2bb6e49)

Downloads

• macOS: .dmg (Apple Silicon)
• Linux: .deb / .AppImage
• Windows: .exe installer

Docker

docker run -d --name attu -p 3000:3000 -v attu-data:/data zilliz/attu:v3.0.0-beta.4

Attu v3.0.0-beta.3

🐛 Bug Fixes

• Fix auto-update restart on macOS — Squirrel.Mac silently hangs on ad-hoc signed builds due to code signature validation failure. Added a manual update fallback that extracts the downloaded zip and replaces the .app bundle when Squirrel fails, ensuring restart always works regardless of code signing status.

📦 Infrastructure

• Improve app release changelog generation

Downloads

• macOS: .dmg (Apple Silicon / Intel)
• Linux: .deb / .AppImage
• Windows: .exe installer

Docker

docker run -d --name attu -p 3000:3000 -v attu-data:/data zilliz/attu:v3.0.0-beta.3

Attu v3.0.0-beta.2

✨ New Features

• Scoped user access management — manage user permissions per database and collection
• Redesigned import dialog — schema guide, file picker, and collection links
• Resizable tree panel — drag to resize the explorer sidebar
• Create collection from empty state — quick-add button when a database has no collections
• Unified context menus — shared menu items across tree and panel views
• Deep link support — attu:// protocol handler for Electron
• Native task notifications — OS-level notifications on task completion
• Sidebar improvements — resource links, reorganized bottom section, persistent expand/collapse state
• Configuration UX — smart value formatting and filters
• Remove collection templates — simplified create collection dialog

🐛 Bug Fixes

• Fix hydration mismatch by deferring localStorage/DOM reads to useEffect
• Fix total count not updating after deleting data entries
• Auto-poll load state after async collection load
• Fix struct array vector index using wrong field path and metrics
• Remove double border between search and table in segments view
• Add missing card background to table tabs in explorer
• Hide duplicate option for built-in AI skills
• Improve error handling for milvus-backup process spawn

📦 Dependencies & Infrastructure

• Upgrade @tanstack/react-router and @tanstack/react-start
• Upgrade drizzle-orm to 0.45.2 and vite to 8.0.8
• Upgrade better-sqlite3 and fix native rebuild for Electron 41
• Update Dockerfile for Node 24 and better-sqlite3 12.8.0
• Add milvus-backup binary download to app release workflow

Downloads

• macOS: .dmg (Apple Silicon)
• Linux: .deb / .AppImage
• Windows: .exe installer

Docker

docker run -d --name attu -p 3000:3000 -v attu-data:/data zilliz/attu:v3.0.0-beta.2

Attu 3.0.0-beta.1

First beta release of Attu v3 — a complete rewrite of the Milvus vector database GUI with modern UX and powerful new capabilities.

⚠️ This is a beta release. Please report issues at GitHub Issues.

New Features

AI Agent

• Built-in LLM-powered Milvus management agent with streaming chat
• 50+ tools covering collection management, data operations (query/search/insert/delete), RBAC, system maintenance, and monitoring
• Customizable skill system
• Per-connection independent agent sessions

Prometheus Monitoring Dashboard

• Built-in metrics dashboard with 16+ core metrics and sparkline visualizations
• Time range selector
• Per-connection Prometheus URL configuration

Cluster Topology Visualization

• Node topology graph and table view
• Node details with MixCoord type mapping

Resource Group Management

• Full CRUD interface with configuration forms, detail panels, and summary cards

HTTP API Playground

• Per-connection interactive REST client for testing Milvus HTTP APIs

Configurations & Environments Viewer

• Runtime configuration display with source filter tags
• Environment variable inspection

Data Editing

• Row-level view/edit/upsert operations in data tables
• JSON validation feedback

Milvus Backup Integration

• Native milvus-backup process management

Slow Requests Viewer

• Cluster slow query log page

Segments Viewer

• Segment details with collection name resolution

UI/UX Improvements

• New design system with light/dark themes
• Connection-centric sidebar navigation with draggable width
• Redesigned import dialog with schema guide and file picker
• Scoped user access management for databases and collections
• Unified context menus across tree and panel views
• Smart value formatting in configurations
• Full English and Chinese i18n coverage
• Connection health monitoring with error cards

Desktop App

• Auto-update via GitHub Releases
• Deep link protocol (attu://)
• Native app menu with macOS integration
• Window size/position persistence
• Native notifications for task completion

Downloads

docker run -p 3000:3000 -e MILVUS_URL=http://your-milvus:19530 zilliz/attu:v3.0.0-beta.1

Attu v2.6.5

New Features

• AI-Powered Schema Generation: A new "AI" tab in the Create Collection dialog lets users describe their use case in natural language and get an AI-recommended collection schema. Supports multiple
  AI providers (including DeepSeek) with in-app API key management and model selection.
• File Upload for Schema Prediction: Upload a CSV or JSON file (up to 10MB) in the AI chat panel when creating a collection. The AI analyzes the file structure to automatically detect field types —
  including arrays and vector fields — and suggests an appropriate schema.
• Modify Replica Count for Loaded Collections: In distributed deployments, users can now adjust the replica count of a loaded collection directly from the collection header without having to
  release and reload it.
• Default Timezone Set to UTC: Database and collection creation now defaults to UTC timezone instead of the user's local timezone, ensuring consistent behavior across teams and environments.

Bug Fixes

• Fixed Authentication for Milvus System Views: System monitoring pages (Tasks, Segments, Replicas, Slow Queries, etc.) now correctly pass auth credentials when connected to a Milvus instance with
  authorization enabled, resolving 401 errors.
• Fixed Task Grid Row Rendering: The Tasks view (Build Index, Compaction, Import, etc.) previously had duplicate React keys causing jumbled or incorrect row display. Each task grid now uses the
  correct task-specific identifier.
• Fixed _default Partition Name Display: The default partition now correctly displays as _default, matching the actual name in Milvus APIs (previously shown as "Default").
• Playground Security Hardening: Patched SSRF and DNS rebinding vulnerabilities in the Playground feature to prevent unauthorized requests to internal network resources.

Attu v2.6.4

This release introduces several new features including AI embedding generation and NGRAM index support, alongside significant performance improvements for handling large numbers of databases and collections. It also includes various bug fixes for timezone handling, UI responsiveness, and data import/export.

🚀 New Features

• AI Embedding Support: Added AI embedding generation for vector search fields and removed restrictions on embedding model selection.
• NGRAM Index: Support for creating NGRAM indexes on VarChar and JSON fields.
• Database Management: Enhanced the Home page with database sorting functionality and improved UI management.
• Nginx Support: Added configuration files and scripts for specific Nginx deployment scenarios.

⚡ Performance Improvements

• Scalability: Optimized listDatabase to gracefully handle a large number of databases.
• UI Responsiveness: Improved rendering performance for the Databases page and Collection creation form by reducing unnecessary re-renders and stabilizing state updates.
• Data Refresh: Unified batch collection refresh logic to prevent redundant API calls.

🐛 Bug Fixes

• Timezone Handling: Multiple fixes to ensure correct timezone persistence, default values (UTC), and correct display logic across database properties.
• Authentication: Fixed an issue where auth form input state was lost when switching tabs.
• Data Import/Export:
  • Supported both schema and simplified formats for JSON imports.
  • Removed db_name from downloaded schema JSON for better portability.
  • Corrected binary vector generation and validation.
• UI/UX:
  • Resolved issues with database switching and collection count display.
  • Fixed URL formatting when clicking the Attu logo.
  • Disabled window shadow on macOS Tahoe to prevent GPU load issues.
  • Improved partition selection and deletion interactions.
• Query & Search: Fixed quoting for VarChar IDs in node search query expressions.
• Schema: Prevented NGRAM index creation on Struct fields where it isn't supported.

🛠 Miscellaneous

• Dependencies: Bumped mdast-util-to-hast to 13.2.1.

Downloads

• macOS: Download the .dmg file for macOS
• Linux: Download the .deb file for Linux
• Windows: Download the .exe installer for Windows