AgentLens adds hash-chained audit trails and OpenTelemetry ingestion for AI agents

·2 views

AgentLens is a young, MIT-licensed open-source project that aims to make AI agent activity tamper-evident. Each event in its append-only journal stores the SHA-256 hash of the previous event, so an edit, deletion or swap breaks verification at that point. The project can also ingest OpenTelemetry GenAI traces on port 3400 and offers an MCP package so agents can log events directly. The author notes that the hash chain is not a cryptographic signature, since someone controlling the database could rebuild it, and that MCP-based logging depends on the agent choosing to call the tool. The reviewer did not deploy the system and reports the demo output as advertised rather than observed.

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)

Supabase 'infinite recursion' error from circular RLS policies, and how to fix it

A common Supabase problem appears when adding sharing, teams, or admin roles: pages that previously loaded return an 'infinite recursion detected in policy' error with code 42P17. The cause is row-level security policies that query the same table they protect, or two tables whose policies reference each other, forcing PostgreSQL into an endless evaluation loop. The author tested this against PostgreSQL 16.15 and notes the error can name the wrong table and may not reproduce in the SQL Editor, which bypasses policies for the table owner. The recommended fix is moving the lookup logic into a SECURITY DEFINER function in a private schema, which runs with privileges that skip the recursive policy check. Example functions and rewritten policies are provided for both the single-table admin case and the two-table sharing case.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Power BI Dashboards Help Sports Teams Turn Athlete Data Into Training Decisions

Sports teams, coaches, and fitness professionals are increasingly using analytics to assess athletes and refine training strategies. Every session and competition produces data on speed, stamina, workload, and recovery, which is hard to analyse manually. Microsoft Power BI can convert this raw information into interactive dashboards that reveal performance patterns. Such dashboards can track progress over time, monitor workload and injury history, and track health and recovery indicators. Experts note that these tools support monitoring and analysis, while medical diagnoses and return-to-play decisions remain the responsibility of qualified healthcare professionals.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Building a Safe Expression Parser to Replace eval Under a Strict CSP

A developer building a live graphing calculator could not use eval because the site's Content-Security-Policy forbids unsafe-eval, a restriction that also shields the site from XSS via user input. Instead, they wrote a roughly eighty-line expression parser that runs in four stages: tokenizing, inserting implicit multiplication, converting to reverse Polish notation via the shunting-yard method, and evaluating. The article highlights three common pitfalls: treating unary minus as ordinary subtraction (so -3^2 must equal -9), requiring right-associativity for exponentiation (so 2^3^2 equals 512), and recognizing implicit multiplication such as 2x without mistaking function names like sin(x) for variables. Parsing once and evaluating the compiled expression at each sample point keeps the plotting loop fast.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Cómo medir si un proceso mejoró de verdad, con cuatro indicadores clave

Un proceso solo puede considerarse mejorado si un indicador definido antes del cambio se mueve en la dirección esperada, sin que otros empeoren, y si ese resultado se mantiene con el tiempo. El artículo recomienda cuatro métricas para pymes: tiempo de ciclo, medido con la mediana y el percentil 90; retrabajo o errores como porcentaje de casos; pasos manuales por caso; y costo por transacción. Advierte que sin una línea base registrada, la supuesta mejora es solo una impresión del equipo. También señala que instalar una herramienta o recibir buenas opiniones no equivale a un resultado. Por último, subraya la importancia de contemplar el volumen de trabajo para interpretar correctamente cualquier variación.

0 comments Read more at DEV Community