AWS CloudWatch EMF Lets Node.js AI Services Emit Custom Metrics via JSON Logs

·3 views

AWS CloudWatch Embedded Metrics Format (EMF) allows developers to embed metric data directly inside CloudWatch log events, eliminating the need for separate PutMetricData API calls. By writing a specially structured JSON log line, engineers can simultaneously capture a log entry and a queryable CloudWatch metric with no added network latency or extra API cost. EMF is particularly useful for high-throughput AI inference endpoints where tracking per-request latency, token counts, and error rates is critical. A valid EMF log must include a top-level @aws block containing a timestamp, namespace, and metric definitions, while additional keys serve as dimensions or measurements. One prerequisite is configuring a CloudWatch Logs subscription filter on the relevant log group so that EMF data is forwarded to the metrics processing pipeline.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

EU Releases Framework to Help Schools Evaluate Digital Education Content

The European Commission published new guidelines on December 16, 2025, to help teachers, school leaders, and education authorities make more consistent decisions when selecting, creating, adapting, and evaluating digital learning resources. Developed by the EU Digital Education Content Expert Group under the Digital Education Action Plan, the guidelines involved over 280 specialists alongside educators, researchers, EdTech providers, and policymakers. The framework introduces a set of quality criteria covering curriculum alignment, learning impact, accessibility, legal compliance, security, technical compatibility, and financial sustainability. Rather than offering a single approval label, the guidance walks educators through the full lifecycle of a digital resource. The document is one of four guideline sets released under the same initiative, with the others addressing ethical AI use in teaching, digital literacy, and informatics education.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer ships unwinnable game level after bug silently skips victory condition

A developer released a wilderness survival game with eight scenarios, only to discover that the hardest paid level was impossible to complete due to a silent code bug. The flaw lay in a timing conflict between two functions: travel() marked hidden locations as discovered before the player arrived, causing arrive() to skip first-visit effects entirely — including the win condition. As a result, reaching the game's designated victory tile in the highland_mine scenario produced no outcome, leaving players stranded in what the game internally recognised as the finish line. All existing validation tests passed because they checked structural correctness — whether locations existed and were reachable — rather than simulating actual gameplay execution. The developer ultimately fixed the bug by reassigning location discovery solely to arrive(), and built a dedicated playthrough harness to catch similar execution-level failures in future.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Storm Reply Builds Automation Toolkit to Slash Manual Work in AWS Architecture Reviews

Storm Reply developed the WAFR Automation Toolkit to address the time and consistency challenges faced during AWS Well-Architected Framework Reviews (WAFRs). A large enterprise client's multi-service AWS environment required a full review spanning six pillars, over 60 questions, and hundreds of best-practice checks, which was taking consultants several days of largely repetitive manual work. The review uncovered critical gaps including excessive IAM permissions, single points of failure, poor cost governance, and incomplete data encryption in transit. Beyond the time burden, manual reviews suffered from inconsistent interpretation of outputs and uneven report quality depending on the individual consultant. The toolkit automates API-checkable evidence gathering using AWS CLI, Prowler, and Steampipe commands, allowing consultants to focus on higher-judgement decisions rather than routine data collection.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Postgres Row-Level Security Has Silent Failure Modes That Green Tests Can Miss

A developer publishing findings on DEV Community discovered that Postgres row-level security (RLS) tests can pass as green even on databases with no active policy, if the test fixture contains only one tenant's data. The investigation found that ENABLE ROW LEVEL SECURITY exempts the table owner by default, while FORCE ROW LEVEL SECURITY does not — a distinction that can mask broken isolation. Testing revealed that reused database connections can silently return zero rows instead of raising an error when tenant context is lost, making failures invisible until a customer notices. A seemingly safe fix — using NULLIF to tolerate empty strings in the policy — was shown to convert loud errors into silent data leaks. The author concludes that a query which raises an exception is preferable to one that quietly returns nothing, as silent failures are far harder to detect in production.

0 comments Read more at DEV Community