dstack 0.19.7__tar.gz → 0.19.8__tar.gz

{dstack-0.19.7 → dstack-0.19.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dstack
-Version: 0.19.7
+Version: 0.19.8
 Summary: dstack is an open-source orchestration engine for running AI workloads on any cloud or on-premises.
 Project-URL: Homepage, https://dstack.ai
 Project-URL: Source, https://github.com/dstackai/dstack
@@ -49,6 +49,7 @@ Requires-Dist: asyncpg; extra == 'all'
 Requires-Dist: azure-identity>=1.12.0; extra == 'all'
 Requires-Dist: azure-mgmt-authorization>=3.0.0; extra == 'all'
 Requires-Dist: azure-mgmt-compute>=29.1.0; extra == 'all'
+Requires-Dist: azure-mgmt-msi>=7.0.0; extra == 'all'
 Requires-Dist: azure-mgmt-network<28.0.0,>=23.0.0; extra == 'all'
 Requires-Dist: azure-mgmt-resource>=22.0.0; extra == 'all'
 Requires-Dist: azure-mgmt-subscription>=3.1.1; extra == 'all'
@@ -116,6 +117,7 @@ Requires-Dist: asyncpg; extra == 'azure'
 Requires-Dist: azure-identity>=1.12.0; extra == 'azure'
 Requires-Dist: azure-mgmt-authorization>=3.0.0; extra == 'azure'
 Requires-Dist: azure-mgmt-compute>=29.1.0; extra == 'azure'
+Requires-Dist: azure-mgmt-msi>=7.0.0; extra == 'azure'
 Requires-Dist: azure-mgmt-network<28.0.0,>=23.0.0; extra == 'azure'
 Requires-Dist: azure-mgmt-resource>=22.0.0; extra == 'azure'
 Requires-Dist: azure-mgmt-subscription>=3.1.1; extra == 'azure'

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/archive/ambassador-program.md

@@ -3,7 +3,7 @@ title: "Get involved as a community ambassador"
 date: 2024-12-18
 description: "Join dstack as an ambassador to grow the community, share knowledge, and help others use dstack."  
 slug: ambassador-program
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/ambassador-program.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/ambassador-program.png
 categories:
   - Community
 ---
@@ -23,7 +23,7 @@ of the `dstack` community, and play a key role in advancing the open AI ecosyste
 [//]: # (Mention:)
 [//]: # (Who we are looking for)
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/ambassador-program.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/ambassador-program.png" width="630"/>
 
 <!-- more -->
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/amd-mi300x-inference-benchmark.md

@@ -3,7 +3,7 @@ title: "Benchmarking Llama 3.1 405B on 8x AMD MI300X GPUs"
 date: 2024-10-09
 description: "Exploring how the inference performance of Llama 3.1 405B varies on 8x AMD MI300X GPUs across vLLM and TGI backends in different use cases."  
 slug: amd-mi300x-inference-benchmark
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-hotaisle-amd-mi300x-prompt-v5.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-hotaisle-amd-mi300x-prompt-v5.png
 categories:
   - AMD
   - Benchmarks
@@ -16,7 +16,7 @@ so we saw this as a great chance to test our integration by benchmarking AMD GPU
 [Hot Aisle :material-arrow-top-right-thin:{ .external }](https://hotaisle.xyz/){:target="_blank"}, who build top-tier 
 bare metal compute for AMD GPUs, kindly provided the hardware for the benchmark.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-hotaisle-amd-mi300x-prompt-v5.png?raw=true" width="750" />
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-hotaisle-amd-mi300x-prompt-v5.png" width="750" />
 
 <!-- more -->
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/amd-on-runpod.md

@@ -106,7 +106,7 @@ cloud resources and run the configuration.
     If you specify `model` when running a service, `dstack` will automatically register the model on
     an OpenAI-compatible endpoint and allow you to use it for chat via the control plane UI.
     
-    <img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-control-plane-model-llama31.png?raw=true" width="750px" />
+    <img src="https://dstack.ai/static-assets/static-assets/images/dstack-control-plane-model-llama31.png" width="750px" />
 
 ## What's next?
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/amd-on-tensorwave.md

@@ -3,7 +3,7 @@ title: Using SSH fleets with TensorWave's private AMD cloud
 date: 2025-03-11
 description: "This tutorial walks you through how dstack can be used with TensorWave's private AMD cloud using SSH fleets."  
 slug: amd-on-tensorwave
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-tensorwave-v2.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-tensorwave-v2.png
 categories:
   - AMD
   - SSH fleets
@@ -18,7 +18,7 @@ In this tutorial, we’ll walk you through how `dstack` can be used with
 [TensorWave :material-arrow-top-right-thin:{ .external }](https://tensorwave.com/){:target="_blank"} using 
 [SSH fleets](../../docs/concepts/fleets.md#ssh).
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-tensorwave-v2.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-tensorwave-v2.png" width="630"/>
 
 <!-- more -->
 
@@ -28,7 +28,7 @@ training and inference.
 Before following this tutorial, ensure you have access to a cluster. You’ll see the cluster and its nodes in your
 TensorWave dashboard.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-tensorwave-ui.png?raw=true" width="750"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-tensorwave-ui.png" width="750"/>
 
 ## Creating a fleet
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/beyond-kubernetes-2024-recap-and-whats-ahead.md

@@ -3,7 +3,7 @@ title: "Beyond Kubernetes: 2024 recap and what's next for AI infra"
 date: 2024-12-10
 description: "Reflecting on key milestones from 2024, and looking ahead to the next steps in simplifying AI infrastructure orchestration."  
 slug: beyond-kubernetes-2024-recap-and-whats-ahead
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/beyond-kubernetes-2024-recap-and-whats-ahead.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/beyond-kubernetes-2024-recap-and-whats-ahead.png
 categories:
   - AMD
   - NVIDIA

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/cursor.md

@@ -3,7 +3,7 @@ title: "Accessing dev environments with Cursor"
 date: 2025-03-31
 description: "TBA"
 slug: cursor
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-cursor-v2.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-cursor-v2.png
 categories:
   - Dev environments
 ---
@@ -17,7 +17,7 @@ Previously, support was limited to VS Code. However, as developers rely on a var
 we’ve expanded compatibility. With this update, dev environments now offer effortless access for users of 
 [Cursor :material-arrow-top-right-thin:{ .external }](https://www.cursor.com/){:target="_blank"}.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-cursor-v2.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-cursor-v2.png" width="630"/>
 
 <!-- more -->
 
@@ -73,7 +73,7 @@ To open in Cursor, use this link:
 Clicking the provided URL will prompt your desktop Cursor IDE to automatically connect to the remote machine via the SSH
 tunnel created by the `dstack apply` command, allowing you to securely work with your dev environment.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-cursor-ide.png?raw=true" width="800"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-cursor-ide.png" width="800"/>
 
 Using Cursor over VS Code offers multiple benefits, particularly when it comes to integrated AI coding assistance and
 enhanced developer experience.

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/docker-inside-containers.md

@@ -2,7 +2,7 @@
 title: "Using Docker and Docker Compose inside GPU-enabled containers"
 date: 2024-10-30
 description: "The latest release of dstack allows for the direct use of Docker and Docker Compose within run configurations."
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-docker-inside-containers.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-docker-inside-containers.png
 slug: docker-inside-containers
 ---
 
@@ -82,7 +82,7 @@ One of the obvious use cases for this feature is when you need to use Docker Com
 For example, the Hugging Face Chat UI requires a MongoDB database, so using Docker Compose to run it is 
 the easiest way:
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-docker-compose-terminal.png?raw=true" width="750"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-docker-compose-terminal.png" width="750"/>
 
 ### docker build
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/dstack-metrics.md

@@ -3,7 +3,7 @@ title: "Monitoring essential GPU metrics via CLI"
 date: 2024-10-22
 description: "dstack introduces a new CLI command (and API) for monitoring container metrics, incl. GPU usage for NVIDIA, AMD, and other accelerators."  
 slug: dstack-metrics
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-stats-v2.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-stats-v2.png
 categories:
   - AMD
   - NVIDIA
@@ -18,7 +18,7 @@ While it's possible to use third-party monitoring tools with `dstack`, it is oft
 track metrics out of the box. That's why, with the latest release, `dstack` introduced [`dstack stats`](../../docs/reference/cli/dstack/metrics.md), a new CLI (and API)
 for monitoring container metrics, including GPU usage for `NVIDIA`, `AMD`, and other accelerators.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-stats-v2.png?raw=true" width="725"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-stats-v2.png" width="725"/>
 
 <!-- more -->
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/efa.md

@@ -3,7 +3,7 @@ title: Efficient distributed training with AWS EFA
 date: 2025-02-20
 description: "The latest release of dstack allows you to use AWS EFA for your distributed training tasks."  
 slug: efa
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/distributed-training-with-aws-efa-v2.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/distributed-training-with-aws-efa-v2.png
 categories:
   - Cloud fleets
 ---
@@ -16,7 +16,7 @@ distributed training workloads across multiple GPUs and instances.
 
 With the latest release of `dstack`, you can now leverage AWS EFA to supercharge your distributed training tasks.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/distributed-training-with-aws-efa-v2.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/distributed-training-with-aws-efa-v2.png" width="630"/>
 
 <!-- more -->
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/gpu-blocks-and-proxy-jump.md

@@ -3,7 +3,7 @@ title: Introducing GPU blocks and proxy jump for SSH fleets
 date: 2025-02-18
 description: "TBA"  
 slug: gpu-blocks-and-proxy-jump
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/data-centers-and-private-clouds.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/data-centers-and-private-clouds.png
 categories:
   - SSH fleets
 ---
@@ -21,7 +21,7 @@ Originally, `dstack` was focused on public clouds. With the new release, `dstack
 extends support to data centers and private clouds, offering a simpler, AI-native solution that replaces Kubernetes and
 Slurm.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/data-centers-and-private-clouds.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/data-centers-and-private-clouds.png" width="630"/>
 
 <!-- more -->
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/h100-mi300x-inference-benchmark.md

@@ -3,7 +3,7 @@ title: "Exploring inference memory saturation effect: H100 vs MI300x"
 date: 2024-12-05
 description: "This benchmark explores how GPU memory saturation affects LLM inference performance and cost, comparing NVIDIA H100 and AMD MI300x."
 slug: h100-mi300x-inference-benchmark
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-v2.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-v2.png
 categories:
   - Benchmarks
   - AMD
@@ -21,7 +21,7 @@ replica on 4xMI300x and 8xMI300x
 
 Finally, we extrapolate performance projections for upcoming GPUs like NVIDIA H200, B200, and AMD MI325x, MI350x.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-v2.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-v2.png" width="630"/>
 
 This benchmark is made possible through the generous support of our friends at
 [Hot Aisle :material-arrow-top-right-thin:{ .external }](https://hotaisle.xyz/){:target="_blank"} and 
@@ -58,7 +58,7 @@ who provided high-end hardware.
 
 ### Cost per token
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-cpt.png?raw=true" width="750">
+<img src="https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-cpt.png" width="750">
 
 As prompt and batch sizes grow, the NVIDIA H100 reaches memory limits, causing a sharp drop in cost-effectiveness. In
 contrast, the 1 FP8 8xMI300x configuration is the most cost-efficient for large prompts.
@@ -66,16 +66,16 @@ contrast, the 1 FP8 8xMI300x configuration is the most cost-efficient for large
 For large prompts, two parallel replicas running on 4xMI300x lose their cost advantage compared to a single replica on
 8xMI300x. The latter offers 51% more memory for the KV cache, improving throughput and reducing cost per token.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-online-requests.png?raw=true" width="750">
+<img src="https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-online-requests.png" width="750">
 
 While 4xMI300x is a cost-effective alternative to 8xH100 for smaller load profiles, it underperforms in online serving.
 8xH100 SXM5 processes 74% more requests per second and reduces TTFT by at least 50% at all QPS levels.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-online-ttft.png?raw=true" width="750">
+<img src="https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-online-ttft.png" width="750">
 
 ### Throughput
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-throughput.png?raw=true" width="750">
+<img src="https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-throughput.png" width="750">
 
 With large prompts and batch sizes, two replicas on 4xMI300x GPUs hit memory saturation when total tokens (prompt
 length x batch size) exceed the available memory for the KV cache. This forces the inference engine to compute KV
@@ -94,12 +94,12 @@ setup, an 8xH200 would process around 2,186 tokens per second (3.4 × 643), thou
 
 #### Replicas on 4xMi300x
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-throughput-2048.png?raw=true" width="750">
+<img src="https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-throughput-2048.png" width="750">
 
 Running two replicas on 4xMI300x delivers better throughput for small to medium prompts than a single replica on
 8xMI300x. 
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-throughput-32784.png?raw=true" width="750">
+<img src="https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-throughput-32784.png" width="750">
 
 This boost comes from distributing the Llama 3.1 405B model across four GPUs, enabling parallel execution. For
 small prompts, a single replica underutilizes the GPUs. Running two replicas doubles the batch size, improving GPU
@@ -107,7 +107,7 @@ utilization and efficiency.
 
 ### Time To First Token
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-ttft-qps-1000.png?raw=true" width="750">
+<img src="https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-ttft-qps-1000.png" width="750">
 
 The 4xMI300x setup provides 768 GB of memory (4 GPUs × 192 GB each), compared to 640 GB with 8xH100 (8 GPUs × 80 GB
 each). However, at 1000 QPS, TTFT for 4xMI300x is over twice as long as for 8xH100
@@ -119,7 +119,7 @@ Despite offering more memory, 4xMI300x lacks the parallelism of 8xH100, leading
 
 ### Time to Serve 1 Request
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h100-mi300x-inference-benchmark-time-1-request.png?raw=true" width="750">
+<img src="https://dstack.ai/static-assets/static-assets/images/h100-mi300x-inference-benchmark-time-1-request.png" width="750">
 
 Processing a single large prompt request with 8xMI300x takes around 11.25 seconds. This latency is mainly due to
 computational demands during the prefill phase, where KV tensors are computed.

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/h200-mi300x-deepskeek-benchmark.md

@@ -3,7 +3,7 @@ title: "DeepSeek R1 inference performance: MI300X vs. H200"
 date: 2025-03-18
 description: "TBA"
 slug: h200-mi300x-deepskeek-benchmark
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/h200-mi300x-deepskeek-benchmark-v2.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/h200-mi300x-deepskeek-benchmark-v2.png
 categories:
   - Benchmarks
   - AMD
@@ -20,7 +20,7 @@ In this benchmark, we evaluate the performance of three inference backends—SGL
 configurations: 8x NVIDIA H200 and 8x AMD MI300X. Our goal is to compare throughput, latency, and overall efficiency to
 determine the optimal backend and hardware pairing for DeepSeek-R1's demanding requirements.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/h200-mi300x-deepskeek-benchmark-v2.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/h200-mi300x-deepskeek-benchmark-v2.png" width="630"/>
 
 This benchmark was made possible through the generous support of our partners at
 [Vultr :material-arrow-top-right-thin:{ .external }](https://www.vultr.com/){:target="_blank"} and 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/inactivity-duration.md

@@ -3,7 +3,7 @@ title: Auto-shutdown for inactive dev environments—no idle GPUs
 date: 2025-02-19
 description: "dstack introduces a new feature that automatically detects and shuts down inactive dev environments, helping you avoid wasted GPU costs."  
 slug: inactivity-duration
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/inactive-dev-environments-auto-shutdown.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/inactive-dev-environments-auto-shutdown.png
 categories:
   - Dev environments
 ---
@@ -18,7 +18,7 @@ a container that has GPU access.
 One issue with dev environments is forgetting to stop them or closing your laptop, leaving the GPU idle and costly. With
 our latest update, `dstack` now detects inactive environments and automatically shuts them down, saving you money.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/inactive-dev-environments-auto-shutdown.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/inactive-dev-environments-auto-shutdown.png" width="630"/>
 
 <!-- more -->
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/instance-volumes.md

@@ -2,7 +2,7 @@
 title: "Introducing instance volumes to persist data on instances"
 date: 2024-11-05
 description: "To simplify caching across runs and the use of NFS, we introduce a new volume type that persists data on the instance."
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-instance-volumes.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-instance-volumes.png
 slug: instance-volumes
 categories:
   - Volumes 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/intel-gaudi.md

@@ -3,7 +3,7 @@ title: Supporting Intel Gaudi AI accelerators with SSH fleets
 date: 2025-02-21
 description: "dstack now supports Intel Gaudi accelerators with SSH fleets, simplifying container orchestration across private clouds and data centers."  
 slug: intel-gaudi
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-intel-gaudi-and-intel-tiber-cloud.png-v2?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-intel-gaudi-and-intel-tiber-cloud.png-v2
 categories:
   - Intel Gaudi
   - SSH fleets
@@ -17,7 +17,7 @@ just leading cloud providers and on-prem environments but also a wide range of a
 With our latest release, we’re adding support
 for Intel Gaudi AI Accelerator and launching a new partnership with Intel.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-intel-gaudi-and-intel-tiber-cloud-v2.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-intel-gaudi-and-intel-tiber-cloud-v2.png" width="630"/>
 
 <!-- more -->
 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/metrics-ui.md

@@ -3,7 +3,7 @@ title: "Built-in UI for monitoring essential GPU metrics"
 date: 2025-04-03
 description: "TBA"
 slug: metrics-ui
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-metrics-ui-v3-min.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-metrics-ui-v3-min.png
 categories:
   - Monitoring
   - AMD
@@ -16,7 +16,7 @@ AI workloads generate vast amounts of metrics, making it essential to have effic
 update introduced the ability to export available metrics to Prometheus for maximum flexibility, there are times when
 users need to quickly access essential metrics without the need to switch to an external tool.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-metrics-ui-v3-min.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-metrics-ui-v3-min.png" width="630"/>
 
 Previously, we introduced a [CLI command](dstack-metrics.md) that allows users to view essential GPU metrics for both NVIDIA
 and AMD hardware. Now, with this latest update, we’re excited to announce the addition of a built-in dashboard within
@@ -27,7 +27,7 @@ the `dstack` control plane.
 The new feature provides an easy-to-use interface for tracking the most essential GPU metrics
 directly from the control plane, streamlining the real-time monitoring process without needing any additional tools.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-metrics-ui-dashboard.png?raw=true" width="800">
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-metrics-ui-dashboard.png" width="800">
 
 Additionally, we’ve renamed the CLI command previously known as `dstack stats` to `dstack metrics` for consistency.
 
@@ -55,6 +55,6 @@ For persistent storage and long-term access to metrics, we still recommend setti
 metrics from `dstack`.
 
 !!! info "What's next?"
-    1. See [Metrics](../../docs/concepts/metrics.md)
+    1. See [Monitoring](../../docs/guides/monitoring.md)
     2. Check [dev environments](../../docs/concepts/dev-environments.md), [tasks](../../docs/concepts/tasks.md), [services](../../docs/concepts/services.md), and [fleets](../../docs/concepts/fleets.md)
     3. Join [Discord :material-arrow-top-right-thin:{ .external }](https://discord.gg/u8SmfwPpMd){:target="_blank"}

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/mpi.md

@@ -3,7 +3,7 @@ title: "Supporting MPI and NCCL/RCCL tests"
 date: 2025-04-02
 description: "TBA"
 slug: mpi
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-mpi-v2.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-mpi-v2.png
 categories:
   - SSH fleets
   - Cloud fleets
@@ -21,7 +21,7 @@ The strength of `dstack` lies in its flexibility. Users can leverage distributed
 system environment variables—such as `DSTACK_NODE_RANK`, `DSTACK_MASTER_NODE_IP`,
 `DSTACK_GPUS_PER_NODE` and [others](../../docs/concepts/tasks.md#system-environment-variables)—to containers.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-mpi-v2.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-mpi-v2.png" width="630"/>
 
 One use case `dstack` hasn’t supported until now is MPI, as it requires a scheduled environment or
 direct SSH connections between containers. Since `mpirun` is essential for running NCCL/RCCL tests—crucial for large-scale

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/nebius.md

@@ -3,7 +3,7 @@ title: Supporting GPU provisioning and orchestration on Nebius
 date: 2025-04-11
 description: "TBA"  
 slug: nebius
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-nebius-v2.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-nebius-v2.png
 categories:
   - Cloud fleets
   - NVIDIA
@@ -16,7 +16,7 @@ developer velocity and efficiency.
 `dstack` is an open-source orchestrator purpose-built for AI infrastructure—offering a lightweight, container-native
 alternative to Kubernetes and Slurm.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-nebius-v2.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-nebius-v2.png" width="630"/>
 
 Today, we’re announcing native integration with [Nebius :material-arrow-top-right-thin:{ .external }](https://nebius.com/){:target="_blank"},
 offering a streamlined developer experience for teams using GPUs for AI workloads.

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/nvidia-and-amd-on-vultr.md

@@ -3,7 +3,7 @@ title: Supporting NVIDIA and AMD accelerators on Vultr
 date: 2025-02-17
 description: "Introducing integration with Vultr: The new integration allows Vultr customers to train and deploy models on both AMD and NVIDIA GPUs."  
 slug: nvidia-and-amd-on-vultr.md
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-vultr.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-vultr.png
 categories:
   - Cloud fleets
   - NVIDIA
@@ -19,7 +19,7 @@ approach.
 Today, we’re excited to share a new integration and partnership
 with [Vultr :material-arrow-top-right-thin:{ .external }](https://www.vultr.com/){:target="_blank"}.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-vultr.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-vultr.png" width="630"/>
 
 This new integration enables Vultr customers to train and deploy models on both AMD
 and NVIDIA GPUs with greater flexibility and efficiency–using `dstack`. 

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/prometheus.md

@@ -3,7 +3,7 @@ title: "Exporting fleet and run metrics to Prometheus"
 date: 2025-04-01
 description: "TBA"
 slug: prometheus
-image: https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-prometheus-v3.png?raw=true
+image: https://dstack.ai/static-assets/static-assets/images/dstack-prometheus-v3.png
 categories:
   - Monitoring
   - NVIDIA
@@ -21,7 +21,7 @@ While `dstack` provides key metrics through its UI and [`dstack metrics`](dstack
 using their own monitoring tools. To support this, we’ve introduced a new endpoint that allows real-time exporting all collected
 metrics—covering fleets and runs—directly to Prometheus.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-prometheus-v3.png?raw=true" width="630"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-prometheus-v3.png" width="630"/>
 
 <!-- more -->
 
@@ -36,7 +36,7 @@ Once Prometheus is set up, it will automatically pull metrics from the `dstack`
 With metrics now in Prometheus, you can use Grafana to create dashboards, whether to monitor all projects at once or
 drill down into specific projects or users.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-prometheus-grafana-dark.png?raw=true" width="800"/>
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-prometheus-grafana-dark.png" width="800"/>
 
 Overall, `dstack` collects three groups of metrics:
 
@@ -46,7 +46,7 @@ Overall, `dstack` collects three groups of metrics:
 | **Runs**   | Run metrics include run counters for each user in each project.                                                                                                   |
 | **Jobs**   | A run consists of one or more jobs, each mapped to a container. Job metrics offer insights into execution time, cost, GPU model, NVIDIA DCGM telemetry, and more. |
 
-For a full list of available metrics and labels, check out [Metrics](../../docs/concepts/metrics.md).
+For a full list of available metrics and labels, check out [Monitoring](../../docs/guides/monitoring.md).
 
 ??? info "NVIDIA"
     NVIDIA DCGM metrics are automatically collected for `aws`, `azure`, `gcp`, and `oci` backends,
@@ -60,7 +60,7 @@ For a full list of available metrics and labels, check out [Metrics](../../docs/
     only accessible through the UI and the [`dstack metrics`](dstack-metrics.md) CLI.
 
 !!! info "What's next?"
-    1. See [Metrics](../../docs/concepts/metrics.md)
+    1. See [Metrics](../../docs/guides/monitoring.md)
     1. Check [dev environments](../../docs/concepts/dev-environments.md),
        [tasks](../../docs/concepts/tasks.md), [services](../../docs/concepts/services.md),
        and [fleets](../../docs/concepts/fleets.md)

{dstack-0.19.7 → dstack-0.19.8}/docs/blog/posts/tpu-on-gcp.md

@@ -126,7 +126,7 @@ and [vLLM :material-arrow-top-right-thin:{ .external }](https://github.com/vllm-
     If you specify `model` when running a service, `dstack` will automatically register the model on
     an OpenAI-compatible endpoint and allow you to use it for chat via the control plane UI.
     
-    <img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-control-plane-model-llama31.png?raw=true" width="750px" />
+    <img src="https://dstack.ai/static-assets/static-assets/images/dstack-control-plane-model-llama31.png" width="750px" />
 
 ### Memory requirements
 

{dstack-0.19.7 → dstack-0.19.8}/docs/docs/concepts/backends.md

@@ -338,6 +338,8 @@ There are two ways to configure Azure: using a client secret or using the defaul
                     "Microsoft.Compute/disks/write",
                     "Microsoft.Compute/disks/read",
                     "Microsoft.Compute/disks/delete",
+                    "Microsoft.ManagedIdentity/userAssignedIdentities/assign/action",
+                    "Microsoft.ManagedIdentity/userAssignedIdentities/read",
                     "Microsoft.Network/networkSecurityGroups/*",
                     "Microsoft.Network/locations/*",
                     "Microsoft.Network/virtualNetworks/*",

{dstack-0.19.7 → dstack-0.19.8}/docs/docs/concepts/dev-environments.md

@@ -175,6 +175,8 @@ name: vscode
 ide: vscode
 
 resources:
+  # 16 or more x86_64 cores
+  cpu: 16..
   # 200GB or more RAM
   memory: 200GB..
   # 4 GPUs from 40GB to 80GB
@@ -187,10 +189,16 @@ resources:
 
 </div>
 
+The `cpu` property also allows you to specify the CPU architecture, `x86` or `arm`. Examples:
+`x86:16` (16 x86-64 cores), `arm:8..` (at least 8 ARM64 cores).
+If the architecture is not specified, `dstack` tries to infer it from the `gpu` specification
+using `x86` as the fallback value.
+
 The `gpu` property allows specifying not only memory size but also GPU vendor, names
 and their quantity. Examples: `nvidia` (one NVIDIA GPU), `A100` (one A100), `A10G,A100` (either A10G or A100),
 `A100:80GB` (one A100 of 80GB), `A100:2` (two A100), `24GB..40GB:2` (two GPUs between 24GB and 40GB),
 `A100:40GB:2` (two A100 GPUs of 40GB).
+If the vendor is not specified, `dstack` tries to infer it from the GPU name using `nvidia` as the fallback value.
 
 ??? info "Google Cloud TPU"
     To use TPUs, specify its architecture via the `gpu` property.

{dstack-0.19.7 → dstack-0.19.8}/docs/docs/concepts/fleets.md

@@ -260,6 +260,10 @@ Define a fleet configuration as a YAML file in your project directory. The file
         2. Hosts with Intel Gaudi accelerators should be pre-installed with [Gaudi software and drivers](https://docs.habana.ai/en/latest/Installation_Guide/Driver_Installation.html#driver-installation).
         This should include the drivers, `hl-smi`, and Habana Container Runtime.
 
+    === "Tenstorrent"
+        2. Hosts with Tenstorrent accelerators should be pre-installed with [Tenstorrent software](https://docs.tenstorrent.com/getting-started/README.html#software-installation).
+        This should include the drivers, `tt-smi`, and HugePages.
+
     3. The user specified should have passwordless `sudo` access.
 
 To create or update the fleet, pass the fleet configuration to [`dstack apply`](../reference/cli/dstack/apply.md):

{dstack-0.19.7 → dstack-0.19.8}/docs/docs/concepts/projects.md

@@ -11,14 +11,14 @@ control which users have access to it.
 In addition to [`~/.dstack/server/config.yml`](../reference/server/config.yml.md), 
 a global admin or a project admin can configure backends on the project settings page.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-projects-project-backends.png?raw=true" width="750px" />
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-projects-project-backends.png" width="750px" />
 
 ## Global admins
 
 A user can be assigned or unassigned a global admin role on the user account settings page. This can only be done by 
 another global admin.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-projects-global-admin.png?raw=true" width="750px" />
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-projects-global-admin.png" width="750px" />
 
 The global admin role allows a user to manage all projects and users.
 
@@ -26,7 +26,7 @@ The global admin role allows a user to manage all projects and users.
 
 A user can be added to a project and assigned or unassigned as a project role on the project settings page.
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-projects-project-admin.png?raw=true" width="750px" />
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-projects-project-admin.png" width="750px" />
 
 ### Project roles
 
@@ -42,7 +42,7 @@ A user can be added to a project and assigned or unassigned as a project role on
 
 Once created, a user is issued a token. This token can be found on the user account settings page. 
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-projects-user-token.png?raw=true" width="750px" />
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-projects-user-token.png" width="750px" />
 
 The token must be used for authentication when logging into the control plane UI
 and when using the CLI or API.
@@ -53,7 +53,7 @@ To use the CLI with a specific project, run the `dstack config` command with the
 
 You can find the command on the project’s settings page:
 
-<img src="https://github.com/dstackai/static-assets/blob/main/static-assets/images/dstack-projects-project-cli.png?raw=true" width="750px" />
+<img src="https://dstack.ai/static-assets/static-assets/images/dstack-projects-project-cli.png" width="750px" />
 
 ??? info "API"
     In addition to the UI, managing projects, users, and user permissions can also be done via the [REST API](../reference/api/rest/index.md).