Skip to main content
Overview
KMS
OS images
Devices
Pre-launch scripts
Init scripts
OS images
dstack 0.5.6
dstack-nvidia-0.5.6
dstack-nvidia-0.5.6
GPU
Feb 6, 2026
0x974a1a9fbb1f1dbe474ad2c77ed30aaf7e6e34eac8173c2a4855f911500a8078
Phala KMS 2026-11
Not allowed
Not allowed
Ethereum KMS
Allowed
Allowed
Base KMS
Allowed
Allowed
Other variants
dstack-0.5.6
dstack-dev-0.5.6
dstack-nvidia-dev-0.5.6
System changelog
Dstack-TEE/meta-dstack@v0.5.6
Changes
Added
guest-agent: Attest API for generating versioned attestations
gateway: WaveKV backend with peer discovery, bootnode support, periodic persistence, and improved cluster orchestration
gateway: multi-domain certificate management with SNI-based resolution, cert configuration UI, ACME account attestation, and configurable DNS TXT TTL/max wait
gateway: multi-port TCP listening via port ranges and deployment script support for multi-port serving
gateway: per-app connection rate limiting
vmm: bridge networking support, DHCP lease PRPC API, and userspace port forwarding
vmm: management APIs UpdateVm and ReloadVms, plus additional metadata in CLI output
vmm-cli: config file support and new update subcommand
vmm-ui: revamped UI (now default), improved layout, device/TEE state display, log follow, git rev display, and dedicated IP UI
guest-agent: systemd socket activation and compatibility socket proxy
kms: auth-simple configuration-based authorization server
sdk: Verifiable Message Signing (Sign/Verify) with signature chain and public key fields
docs: conntrack tuning guide for high-concurrency gateways
docs: bridge networking guide updates and cluster deployment documentation
vmm: OpenAPI documentation output
Changed
gateway: deployment scripts refactored to externalize config and add bootstrap flow
gateway: IP allocation scheme updated for larger address space
gateway: DNS configuration defaults and UI settings refined (TTL, max wait, default port behaviors)
toolchain: Rust pinned to 1.92 and additional no_std target added for CI
attestation: refactored for multi-provider support
vmm: default shared mode set to 9p
dependencies: updated dcap-qvl to 0.3.10 and various dependency bumps (lodash, hono, go-ethereum, tracing-subscriber, etc.)
docs: reorganized and consolidated (confidential AI, verification tutorial, GPU TEE guide, FAQ, SDK docs, main index)
vmm-ui: regenerated and synchronized UI assets
Fixed
vmm: VM config loading issues and multiple UI display bugs
host-api: forbid listening on non-vsock addresses
vmm: trigger port forward reconfiguration on update-ports
runtime: Docker mount socket path compatibility (/run vs /var/run)
runtime: create mount points before rbind mount
sdk/js: isReachable behavior for v0.5.x
gateway: improved error messages for client registration and cert flows
ct_monitor: TLS certificate verification behavior
tooling: clippy warnings, formatting, and CI stability fixes
Security
upgraded dcap-qvl to 0.3.10 to address CVE-2026-22696
Full notes for this version
Products
Solutions
Developers
Resources
About Us
Pricing
Contact
Sign in
Contact
Sign in
Sign up
Products
Confidential VM
GPU TEE
Private AI Models
Model Catalog
dstack
GPU Marketplace
H100
H200
B300
Solutions
AI Agents
Private AI Inference
Private AI Training
Private AI Data
Phala vs AWS Nitro
Phala vs GCP Confidential VM
Phala vs Tinfoil
Financial Services
Healthcare Research
AI SaaS Platform
Decentralized AI
Legal AI
All use cases
Developers
Docs
Quickstart
CLI
Templates
Deploy CVM
Deploy dstack
Run GPU TEE
Verify Attestation
API Reference
Research
GitHub
DeepWiki
Service Status
Resources
Blog
Research
Learn
Tags
By use case
Trust Center
Security
Subprocessors
Startups
Enterprise AI
GPU Providers
Partnerships
About Us
About Phala
Build History
Contact
Wallet, Stake, Bridge
About Phala Network
Careers
Pricing
Contact
Sign in
Sign up