Platforms
Adopt
-
Colima is now our go-to alternative to Docker Desktop on macOS. We continue to use it on several projects to provision the Docker container run time in a Lima VM, to configure the Docker CLI on macOS and to handle port-forwarding and volume mounts. Colima can be configured to run containerd as its run time, which is also the run time on most managed Kubernetes services, improving the important dev-prod parity.
Trial
-
Events are common mechanisms in event-driven architecture or serverless applications. However, producers or cloud providers tend to support them in different forms, which prevents interoperability across platforms and infrastructures. CloudEvents is a specification for describing event data in common formats to provide interoperability across services, platforms and systems. It provides SDKs in multiple languages so you can embed the spec into your application or toolchain. Our teams use it not only for cross-cloud platform purposes but also for domain event specification, among other scenarios. CloudEvents is hosted by the Cloud Native Computing Foundation (CNCF) and now runs as an incubator project that has been gaining increasing industry attention.
-
DataOps.live is a data platform that automates environments in Snowflake. Inspired by DevOps practices, DataOps.live lets you treat the data platform like any other web platform by embracing continuous integration and continuous delivery (CI/CD), automated testing, observability and code management. Our teams are using it for managing the lifecycle of data products that includes development, branching and deployment of both code and data. With its automated environment management, it's very easy to build environments based on feature branches, modify them and destroy them automatically. It’s also worth noting its declarative specification (SOLE) capability, which enables a streamlined developer experience. This allows teams to reduce the time it takes to build data products from months to days. Our teams have been using DataOps.live successfully in production, and that's why we recommend this platform when working with Snowflake.
-
Significant developments have happened in the AI landscape since we first blipped Google Cloud Vertex AI. Since May 2023, Google has introduced several services and features to enrich this realm. These additions include Model Garden, a repository of 100+ pre-trained models; Generative AI Studio, a console intended to rapidly explore and prototype generative AI models; and Vertex AI Extensions which provides fully managed developer tools to connect AI models and real-time data or actions via APIs. The platform has evolved to offer GenAI models and integration support, and we’re excited to use it more extensively.
-
Since we last wrote about Immuta, our teams have gained significant experience with this data security platform. Its highlights include the ability to define subscription and data policies as code, version control and the ability to deploy these policies automatically to higher environments. Its ABAC support allows us to associate tags to data sources; if the same tag is associated with the user, access is granted. By leveraging Immuta and Snowflake integration we've been able to automate granting access to data products or data sets in a self-serve fashion. When the "user" requests access to a data product or a data set, the data product tag is then associated with the "user" as an attribute upon approval. Since the attribute on the "user" matches the tag on the data source, access is granted automatically courtesy of Immuta's Global Subscription policy. It's also worth noting Immuta's data masking policies which preserve data privacy by masking and restricting PII information to a specific user. Additional access to sensitive information at a much more granular level can be defined using row-level security policies that ensure users only have access to the specific data they're authorized to view. We've been happy with Immuta which is why we’re moving it to Trial — it provides a good developer experience and makes it easier for large organizations to manage data policies.
-
Lokalise is a fully automated localization platform that allows for context-specific translations. Our teams use the Lokalise API in their ETL pipelines or development workflows to translate localizable information. Lokalise supports multiple file formats for the localizable strings. One aspect to highlight is the ability to upload an entire file, where each key-value pair is treated as a separate record and translated. Under the hood we leveraged Lokalise's integration with Google MT to take care of the translations. The Lokalise web UI provides ease of access to human reviewers to verify the translations, shorten them and rephrase them as they deem fit. In the past we've highlighted similar tools such as Phrase. Our teams have had a good experience with Lokalise, and we recommend you evaluate the platform for collaborative translation workflows.
-
Orca is a proprietary cloud security platform that identifies, prioritizes and remediates security risks and compliance issues. It supports major cloud providers and hybrid setups. Orca has extensive security queries/rules to continuously monitor deployed workloads for misconfigurations, vulnerabilities and compliance issues. It supports cloud VMs, serverless functions, containers and Kubernetes applications for the deployed workloads. These inbuilt security rules are consistently updated to keep pace with the evolving compliance standards and threat vectors. Since Orca is agentless, it offers a good developer experience and is easy to set up. Another notable feature is that it facilitates shift left security. Our teams use Orca CLI for scanning container images and IaC templates for vulnerabilities and misconfigurations as a pre-commit hook or as part of CI/CD workflows. It also continuously monitors and scans container registries (e.g., AWS ECR) for vulnerable base images or weak OS dependencies for already published images. Based on our teams’ experiences, Orca provides a unified view of the security posture across the path to production, and for that reason we place it in Trial.
-
Trino, previously known as PrestoSQL, is an open-source, distributed SQL query engine designed for interactive analytic queries over big data. It is optimized to run both on-premise and in the cloud. It supports querying data where it lives, including Hive, Cassandra, relational databases and even proprietary data stores. For authentication mechanisms, it supports password-based authentication, LDAP and OAuth. For authorization and access control, Trino provides the ability to grant access at the catalog, schema and table levels. Our teams used resource groups spliced according to consumption patterns like visualization, reporting or machine learning use cases to manage and limit resource usage. The JMX-based monitoring provides a rich set of metrics to enable cost attribution at query or user level. Our teams use Trino as a gateway for data access across a variety of sources. When it comes to querying extremely large-scale data, Trino is a safe bet for our teams. Presto, the Facebook project from which Trino originates, was first featured in the Radar in November 2015.
-
Wiz is another contender in the maturing cloud security platform landscape that allows its users to prevent, detect and respond to security risks and threats in one platform. Wiz can detect and alert on misconfigurations, vulnerabilities and leaked secrets both in artifacts that have yet to be deployed to live environments (container images, infrastructure code) as well as live workloads (containers, VMs and cloud services). It also contextualizes findings to the customer's specific cloud landscape to enable response teams to better understand the issue and prioritize mitigations. Our teams have had good experience with Wiz. They find that Wiz is rapidly evolving and adding new features, and they appreciate that it enables them to detect risks and threats sooner than some other similar tools as it continuously scans for changes.
Assess
-
With the current upheaval in the micro-blogging platform space, the ActivityPub protocol is gaining prominence. ActivityPub is an open protocol for sharing information such as posts, publications and dates. It can be used to implement a social media platform, but the key benefit is that it delivers interoperability between different social media platforms. We expect ActivityPub will play a significant role in this space, but we're mentioning it here because we're intrigued by the possibilities beyond the obvious use cases in social media. An example is ActivityPub support for merge requests, recently proposed for GitLab.
-
Azure Container Apps is a managed Kubernetes namespace as a service that streamlines the deployment of containerized workloads by eliminating the need for intricate maintenance of Kubernetes clusters and underlying infrastructure components, consequently diminishing operational and administrative burdens. However, it’s essential to tread carefully while considering this option; currently in its developmental phase, it has exhibited inconsistencies in the Azure portal's representation of its capabilities and encounters integration hurdles, particularly with the standard Terraform provider for Azure lagging in mirroring the tool's actual functionalities. Given all this, we recommend assessing this tool carefully.
-
With the huge interest in generative AI, many solutions have sprung up to access the major models. If considering or already using Azure, then it's worth assessing Azure OpenAI Service. It provides access to OpenAI's GPT-4, GPT-35-Turbo and Embeddings models through a REST API, a Python SDK and a web-based interface. The models can be adapted to tasks such as content generation, summarization, semantic search and natural language to code translation. Fine-tuning is also available via few-shot learning and the customization of hyperparameters. In comparison to OpenAI's own API, Azure OpenAI Service benefits from Azure's enterprise-grade security and compliance features; it's also available in more regions, although availability is limited for each of the larger geographic regions, and, as of writing, India is not included.
-
There are many emerging large language models (LLMs) in the English-speaking world. Although these models are usually pretrained with multiple languages, their performance in other languages may not be as good as in English. ChatGLM, developed by Tsinghua University, is an open bilingual language model optimized for Chinese conversation based on the General Language Model architecture. Since Chinese can be more complex than English with its different word segmentation and grammar, it’s important to have an LLM optimized for Chinese. Our team found ChatGLM beat other LLMs in accuracy and robustness when we built a Chinese emotion detection application for a call center. Considering many LLMs aren’t available in China due to licensing or regional restrictions, ChatGLM became one of the few open-source options.
-
Chroma is an open-source vector store and embedding database, useful for enhancing applications powered by large language models (LLMs) by facilitating the storage and utilization of domain knowledge in LLMs, which typically lack internal memory. Particularly in text-to-text applications, Chroma can automate the intricate process of generating word embeddings and analyzing similarities between them and query embeddings, thereby considerably streamlining operations. It also gives you the option to store custom embeddings, fostering a blend of automation and customization. In light of its capabilities to enhance the functionality of LLM-powered applications, we advise teams to assess Chroma, tapping into its potential to refine the way domain knowledge is integrated into such applications.
-
UX research platforms such as Dovetail offer organizations a tool to understand and improve their customer experience. With it, businesses are able to quickly and easily gain valuable insights into their customer's needs, preferences and behaviors by collecting and analyzing data from customer feedback, surveys, interviews and more. Sentiment analysis, customer segmentation, market research or data analysis and insight generation are valuable tasks in product development — these match what LLMs are good at, hence we see a great potential for disruption in the product development field.
Kraftful — a self-described copilot for product builders — has taken the lead. It's only in beta and you must provide your email to access the feature list. We've played with it and seen great results. You can plug more than 30 sources of user feedback into the platform and it will analyze the data and identify feature requests, common complaints, what users love about the product and even name your competitors. To gather more details, you can ask questions like you would to ChatGPT or Google Bard — the benefit here is it's optimized for your data. Once you prioritize what will be addressed from the user's feedback, Kraftful generates user stories for you based on all underlying data — including acceptance criteria — making it a great assistant to even very experienced product managers and business analysts.
-
With the rise of Generative AI-powered applications, we see a pattern of storing and efficiently searching embeddings vectors for similarities. pgvector is an open-source vector similarity search extension for PostgreSQL. We quite like it because it enables us to search the embeddings in PostgreSQL without moving the data to another store just for similarity search. Although there are several specialized vector search engines, we want you to assess pgvector.
-
Pinecone is a fully managed, developer-friendly and cloud-native vector database with a simple API and no infrastructure hassles. Pinecone serves filtered query results with low latency at the scale of billions of vectors. Our teams have found vendor databases and Pinecone in particular very helpful and quick to get started for use cases like storing a team’s knowledge base or help desk portal content rather than fine-tuning complex LLMs.
-
wazero is a zero dependency WebAssembly (WASM) run time written in Go. Although the run time itself is language neutral, we wanted to highlight wazero for Go developers because it offers a convenient way to extend your Go programs with WASM modules written in any conformant languages. It has no dependency on CGO, so you can easily cross compile your Go applications to other platforms. Although you have a choice when it comes to run times for WASM, we think wazero is worth assessing.
Hold
- New
- Moved in/out
- No change
Unable to find something you expected to see?
Each edition of the Radar features blips reflecting what we came across during the previous six months. We might have covered what you are looking for on a previous Radar already. We sometimes cull things just because there are too many to talk about. A blip might also be missing because the Radar reflects our experience, it is not based on a comprehensive market analysis.
Unable to find something you expected to see?
Each edition of the Radar features blips reflecting what we came across during the previous six months. We might have covered what you are looking for on a previous Radar already. We sometimes cull things just because there are too many to talk about. A blip might also be missing because the Radar reflects our experience, it is not based on a comprehensive market analysis.
