This feature makes it easier for you to access your existing Artifacts. With this update, you can mount your Artifacts directly to your build instances and access files from your buckets without downloading them!

You can add multiple artifacts to the build instance by going to the instance settings and under artifact mounting. The data is only downloaded when you or a program reads the file.

The mounted Artifacts behave as standard directories, so anything you could do with a regular file or directory is directly applicable here too. We've integrated the Artifacts mounting feature now into Projects. Hence, any build instance you create inside your project has the project artifact mounted by default on the path /home/ubuntu/project/bundle-<project-id>. This feature simplifies your life by providing a single place for your data, which is always available whenever you create an instance.

GUI for Jobs

We're also releasing a new GUI for Jobs and Deploy! This upgrade simplifies setting up and deploying code, so you can get your projects up and running faster.

With this update, you can now use a NimbleBox-designed pre-set job template to quickly set up a Job or Deploy. To use this feature, connect your GitHub account to NimbleBox, and select a repository with its branch and the file in the repository you would like to run. You can then choose your requirements file and save. A job is triggered that runs your programmed function.

Improvements & Fixes

Persistent logging for Models
Feature gates support for Models
Supporting GPU choices in Jobs & Deploy including A10G and A100 GPUs
Un-mounting artifacts when instance is stopped
Added project logs when artifact is deleted
Added auto shutdown notification
Deploy: access key for each model
Deploy: new model states
Deploy: updated log streaming for models, added pagination

Version

3.4.0

April 4, 2023

Traffic Distribution

Testing multiple models is critical but challenging when building an ML service. Worry not; we just launched traffic distribution- the precursor to a full-fledged deployment testing dashboard.

With traffic distribution, you can dynamically allocate traffic to different versions of your models based on various criteria. This feature gives you greater control over your ML deployments, allowing you to optimize your services for multiple use cases and freely experiment with your models.

Artifacts Integrations: AWS S3, Azure Blob, and Google Cloud Storage

At NimbleBox, we believe managing data should be easy and efficient, regardless of where it's stored. That's why Artifacts are a crucial component of our platform, designed to help you manage your data and models in a scalable and reproducible way. With Artifacts, you can store, version, and share your datasets, models, and other artifacts with your team, making collaborating on ML projects, reproducing experiments, and scaling your workflows easier.

With this exciting update, you can seamlessly integrate your data stored in AWS S3, Azure Blob, or Google Cloud Storage with Artifacts, allowing you to work with your data across different cloud storage services without any manual transfer between services.

Improvements & Fixes

Create Projects from GitHub and BitBucket repositories
Fixed instance not starting with GPU for non-default regions
APIs for monitoring live deployments
Instance: backup and template bug fix
Integration: prod integration list fix
Project: BYOC project and delete API
Added workspace status update API
Added a networking tab for traffic distribution
Enabled line smoothening in the monitoring dashboard
Added status to templates card
Changed menu order in instance card
Add sorting and animation in the build list
Changed Instance Settings -> Configure
Updated saving & exiting Hardware Choice
Changed the order of tabs in Build
Added project image feature
Created a new onboarding flow
RuleBuilder: Add restrictions in the rule builder input
Build: Update Sort icon
Project: Add Sorting in the projects list
Project: Cards animations added to the projects list
Project: Add search in the projects list
Removed personal workspace details from the workspaces list
Traffic distribution: Slider issue fixed
Fixed update hardware issue in Build
Fixed projects bug where the same tag was getting added twice
Fixed upload to GCS issue in relics

Version

3.3.2

March 7, 2023

Better Build Dashboard with search, sort & filters

Can't find the instance you were working on yesterday? We understand, and well now it's easier than ever to find it.

You can now directly search, sort, or filter your instances using tags such as "Created by me", "Shared with me", and "Running instances". By applying one or more of these options, you can quickly locate the instance you need.

Add Alerts to your Experiments

In a world where it is harder than ever to keep track of failed models, we bring you- Alerts!

You can create alerts of varying complexity by using logical operators such as "and" and "or".

To set up an alert, go to the Experiments section and select the Alert tab. From there, click on "New Rule" and provide a name for the rule. You can then select the logical operation you want to use from the drop-down menus and write a message that will be sent to your team via email whenever a rule is broken.

Added follow-tail to logs

You can now follow the tail in logs! Simply enable the "Follow Trail" toggle located in the top-right corner of the page.

This will cause the screen to automatically scroll down as new logs are generated, allowing you to easily keep track of the most recent logs as they appear.

Naming & re-naming modules

Like any other fast-moving development team, we're not perfect. We named Build Instances as Projects and then realised we have a new feature that could be perhaps more-aptly named Projects. So we are renaming Build Projects to Build Instances.

Sadly, our much-loved internal name for s3 bucket integrations → Relics, is being renamed to Artifacts. It was much harder for beginners in MLOps to relate to the term Relics, and Artifacts seemed a more apt name for the feature.

We're always listening to you, so let us know how you feel about this. (tag NimbleBoxAI on twitter)

Improvements & Fixes

Model deployment with S3 was implemented
Added a tooltip to the sidebar menu when in the collapsed state
Build's instance card is now clickable
Launch Instance UI improvements
Not allowing special characters in Projects tag names
Fixed double scroll issue in Artifacts
Fixed Frontend not showing some logs in Monitoring
Fixed upload file bug in Artifacts
Deploy: A bug was fixed where the resource was not being persisted in the database for models
Datadog plugin was enabled for APISix
A refresh token was added for models,
Jobs: _add_refresh_token method was added
Jobs & Deploy: Circular import was fixed
Jobs & Deploy: Bug fixes in the init container
Jobs & Deploy: Bug fixes in termination helper
Jobs & Deploy: updated vulnerable packages
Implemented the ability to close logs modal on escape
Added Activity logs tab inside deployment
Added "view curl command" for models in a ready state in Deploy
Deploy: status field mask for a model update call
Deploy: unique model_access_key
Deploy: implemented SetModelMetadata FeatureGate
Model Deployment Failure UI bug fixed
Added workspace id to delete experiment API in Monitoring
Deploy: not allowing overriding of access_key for pinned models
Deploy: add auth_info-related environment variables for models
Add access tokens to models in Deploy

Version

3.3.1

February 24, 2023

Job Resources UI

We have introduced a new user interface for viewing and modifying Job Resources.

You can change the CPU, GPU, RAM and storage resources when creating or editing a job, allowing you to fine-tune your job settings and optimize your resource usage.

You can also see the timeout set for a job and modify the max retries if it fails.

Improved Logging

Introducing a new and improved logging system for Jobs and Deployments. Run Logs and Model Logs now have beautiful and intuitive interfaces, making it easier for users to log and analyze their data.

You can now filter and search logs based on severity level and keywords.

Improvements & Fixes

Relics naming issue fixed
Launch Instance UI improvements
Dependency updates to fix vulnerabilities
Switching servers to multiple sidecars for better performance and latencies
Better Internal authentication between our internal services (Relics, LMAO)
BYOC Jobs: Added functionality for [un]pause, reschedule, logs
BYOC Jobs: Fixed job creation issue
Infrastructure upgrades for faster execution of BYOC tasks
Fixed Relics bug: Conflicting name lead to 500 error
Fixed Relics permission issue: Workspace Admins could not modify Relics
Fixed BYOC Templates bug: Deleting an issue
Fixed BYOC Backups bug: Backups listing was incorrect
Fixed Join workspace bug

Version

3.3.0

February 14, 2023

General Availability of Monitoring

Experiment Tracking is the key to successful MLOps.

So with this feature, you can now run multiple experiments in parallel and track any metric, visualizing the results with beautiful charts. In addition, this new tool provides a centralized platform for managing all information related to experiments, making it easier for your research and engineering team to work together and achieve your targets.

Read more about this launch on our blog.

Improved Jobs Scheduling

You can now Schedule Expressions for Job scheduling.

Set a schedule like 45 9 1 * * to schedule the job at "9:45am on the 1st of every month".

You can also see the next 2 jobs that would be scheduled once you save.

This allows you to focus on the task at hand, without having to worry about your job schedules.

Jobs Resources

Now you have the option to configure Kubernetes-style hardware resources for a Job directly from the UI. This feature is available on NimbleBox cloud as well as for your GCP/AWS/Azure cloud integration.

With this feature, you can easily adjust parameters such as CPU and GPU allocation (with the option to even add multiple GPUs), memory usage, disk space, max retries in case of a failed job and network timeout.

To use this feature, you can simply navigate to the Jobs dashboard. Select the job you wish to modify, go to Settings and click on "Edit Resources".

You can use the resource allocation fields and dropdown menus to adjust the desired parameters. Once you've made the changes and saved them, the job will be allocated the newly configured resources for the next execution.

Relics Activity Logs, MIME type support & RBAC

Now Relics comes with activity logs. You can go to your Relic and click on the "Logs" tab to see all relic activity with timestamps.

With MIME type support, you can now see your images and file content directly in the Relics dashboard.

Relics now also comes with Role-Based access control allowing you to choose which team members you want to share your Relics with.

Click on the corresponding Relic and click on the "Manage Access" tab to add or remove access.

BYOC Instance Backup

You can create snapshot backups of the instances on your cloud easily from the Build Dashboard.

Navigate to the "Backup now" option in your instance settings.

Improvements & Fixes

Improved backend for Monitoring Dashboard
Fixed Jobs logs not showing up in some cases
Fixed GPU machine start issue in AWS (BYOC)
Fixed hardware configuration issues in GCP Integration (BYOC)
Fixed the incorrect deployment URLs that were shown in UI for BYOC Deploy
Fixed a bug in the wallet system which led to credits not showing up properly

Version

3.2.0

February 3, 2023

Log streaming & GPU in Deploy

In this release, we've added log streaming to Deploy, both in the UI and nbox, for improved visibility and tracking. In addition, deploy now has GPU compatibility, allowing for more efficient processing. Additionally, you can now specify hardware resources for model deployment.

Phasing out personal workspaces

We have decided to phase out personal workspaces and are saying goodbye to the Personal Pro plan. However, we will continue to support existing plans for the time being. As part of these changes, we have also removed the trial of workspaces and are introducing - a wallet for workspaces. These updates are aimed at streamlining our product offerings.

BYOC Upgrades

BYOC Templates are now supported
BYOC Build now has GPU Compatibility
BYOC Deploy now has GPU Compatibility

Improvements & Fixes

Build Projects are now called Instances
Asking for cookie consent - Blocking user cookies if the user denies consent
Displaying error message on instance card if storage is full
Onboarding changes & bug fixes
Added bottom margin to allow for HubSpot container
Fixed increasing storage in Build
Fixed displaying incorrect storage percentage in Build
Fixed regions not rendering for the "all" option in Build
Model deployment can now be stopped while it's in progress
Disabled model logs when the model is stopped

Version

3.0.0

January 6, 2023

Bring Your Own Cloud (BYOC)

We are excited to introduce our Bring Your Own Cloud (BYOC) feature on the NimbleBox.ai platform. With this new feature, you can now integrate your own private cloud environment on our platform for easy management and scaling. You can use your cloud integration with our existing products- Build, Jobs, Deploy, and Relics.

We have additional security features to ensure your data remains secure and private. We support a wide range of cloud providers, including Amazon Web Services, Microsoft Azure, and Google Cloud Platform. We have added additional tools and resources to help you set up and manage your private cloud environment.

Fixes and Improvements

We implemented the ability for you to create instances in any region
You can now set custom subdomains for your Build instances
BYOC Integrations list has all option to list all instances, jobs, and deployments from all cloud integrations.
We implemented the ability for you to update hardware configurations easily while using BYOC
Display of instance details modified for a better user experience
Improved functionality of the instance list feature
Better logging for Deployments based on errors
We made various minor improvements and enhancements to the system