KPIAssembler Ruby gem uses AI to propose and certify KPIs before publication

·4 views

A developer has released KPIAssembler, an open-source Ruby gem designed to address flawed metric definitions in data pipelines. The tool inspects a live database schema, uses an AI model such as Google Gemini or a local Ollama instance to propose KPI recipes, and then runs deterministic code-level checks before any metric can be published. Certification requires each candidate query to be read-only, reference existing tables, include tenant scoping where applicable, avoid unconstrained JOINs, and survive both an EXPLAIN check and a sample-period replay. Any KPI that fails these checks is marked as a draft with explicit failure reasons rather than being silently rejected. Schema-based heuristics serve as a fallback if the configured LLM is unavailable, ensuring users always receive metric candidates.

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)

LLM Agreement Study Finds Models Echo Their Own Priors, Not Message Content

A small art project encoding sentences as only word-length sequences prompted researchers to test whether large language models could recover intended meaning from that minimal signal. The experiment compared model readings of a true message's length sequence against readings of a different length sequence and against randomly assembled texts, measuring positional agreement across four runs. Results showed that treatment-arm agreement never consistently separated from the prior-control arm across all four runs, suggesting models were converging on shared priors rather than decoding actual message content. An early version of the study contained a methodological flaw where the random-baseline vocabulary was drawn from the control arm's own limited readings, artificially inflating the floor and nearly obscuring the real finding. The corrected conclusion is that when evaluating model self-consistency, the meaningful baseline is the model's own prior output, not random chance.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Chinese CS Student Reflects on Startup Job That Ended in Dismissal

A software engineering student from China, who goes by ACatNight, has shared reflections on a formative internship at a small startup during their second semester. The student worked on real-world projects and bugs but was eventually let go after making a mistake on the job. Despite the setback, the experience reshaped their understanding of software development, highlighting the importance of system design, collaboration, and handling pressure beyond just writing code. After returning to school, the student resumed personal projects and is currently studying C# .NET backend development and AI applications using tools like ASP.NET Core, Docker, and Redis. They are also rebuilding a side project called StudyTree, an AI-powered learning reflection tool designed to help users identify knowledge gaps and plan what to learn next.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Solo Developer Shares How a Missing PostgreSQL GRANT Broke Multi-Tenant CRM in Production

A Brazilian solo full-stack developer built FunilChat AI, a multi-tenant WhatsApp CRM with an AI agent designed to handle sensitive customer documents including SSN and tax ID equivalents. The system uses PostgreSQL row-level security to isolate data across multiple business clients within a single database, avoiding the complexity of per-tenant deployments. Months into production, an n8n automation workflow silently failed to create handoff notifications because a database migration had granted INSERT on the notifications table but omitted the required GRANT USAGE and SELECT on the associated sequence. The missing sequence permission prevented auto-generated IDs from being created, and the fix required a follow-up migration to explicitly grant those sequence privileges. The project has been running continuously since April 2026 and now spans 21 tables under row-level security across release 2.21.0.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Students Can Make the Most of $100 Azure Cloud Credits

Many students who receive Azure cloud credits tend to spend them on random experiments rather than building meaningful projects, according to a guide published on DEV Community. The article argues that the real value of cloud credits lies in using them to develop portfolio-worthy skills, not simply exploring available services. It recommends seven project categories, ranging from personal portfolio websites and REST APIs to containerized applications using Docker. Each project type is mapped to practical skills such as backend development, database design, cloud deployment, and DevOps workflows. The guide emphasizes that deployed, publicly accessible projects demonstrate engineering ability far more effectively than code that only runs locally.

0 comments Read more at DEV Community