As a new Technology Radar release with many interesting techs, a few drew my attention.
Techniques
- Retrieval-augmented generation (RAG): Useful pattern to improve LLM result. A search result from the vector database is combined with the prompt to be sent to LLM, resulting in better quality than just using LLM.
- Automatically generate Backstage entity descriptors.
- Combining traditional NLP with LLMs can be cost-effective.
- Continuous compliance: Integrating security checking tools in the development pipeline sounds beneficial and practical.
- Edge functions: Code execution via CDNs. e.g. Cloudflare Workers can be useful
- Security champions: Having a security champion in a team is a good idea.
- Text to SQL: LLMs can do it, or Coplit in VS Code can help. But a Python RAG framework like Vanna can do especially well with Text to SQL. NSQL and domain-specific LLM are worth looking into.
- Tracking health over debt: An example of an action item is 1. Measure the system health of your organization based on your priority to get the baseline. e.g. REA’s example of check list 2. Dedicate time to improving the system’s health. e.g., automation on testing/CICD, Using a framework, etc. 3. Continuously track it. e.g. quarterly. to see the improvement from the baseline. 4. Cultivated engineering excellence culture.
- The ability to turn natural language into SQL and display the results has always been possible, but LLM advances are improving it. Slack is now able to build, deploy, and troubleshoot in natural language (https://www.promptops.com/devops/), LLM-armed autonomous programs that go beyond chat and execute instructions independently (https://github.com/Significant-Gravitas/AutoGPT), and there are frameworks (https://microsoft.github.io/autogen/) that make this possible, as well as AI (https://www.driverai.com/) that can explain complex legacy code.
Platforms
- CloudEvents: a specification for describing event data in a common format across platforms and systems
- Arm in the cloud: An arm-based process in the cloud, which leads to container technology, can reduce the cost of hosting app services.
- Azure Container App: A simpler way to host containerized apps to Azure compared to Azure Kubernetes Service.
- Aure OpenAI Service: Enterprise-grade OpenAI service in Azure.
- DataHub: It supports data lineage, data discoverability and data governance.
- Infrastructure orchestration platform: e.g. Terragrunt a thin wrapper for Terraform
- Pulumi: Alternative to Terraform.
- Rancher Desktop: Alternative to Docket Desktop, which has not been free for business since 2024.
- Weights & Biases: The AI developer platform.
- Keycloak: Open source identity and Access Management.
(It is not on this radar, but it looks very useful.)
Tools
- 42Crunch API Conformance Scan: Scan the API and report the vulnerability.
- Bruno: an open-source desktop alternative to Postman and Insomnia.
- GitHub Copilot: “You can now also significantly expand the chat context when asking questions about your code by using the
@workspacetag.” - Microsoft SBOM tool: open-source tool to generate SBOM(Software Bill of Materials)
- Pop: pair programming tool.
- Renovate: Alternative to Dependabot
- Terrascan: Terrascan is a static code analyzer for Infrastructure as Code
- Akvorado: network flow collector, enricher and visualizer
- LinearB: Qunatify engineering efficiency.
- Ollama: Open-source to run LLMs on local machines.
- QAnything: Question and Answer based on Anything is a local knowledge base question-answering system designed to support a wide range of file formats and databases, allowing for offline installation and use.
Languages & Frameworks