Programming

ProgrammingDEV Community ·

How to Build a Secure Node.js AI API Gateway With Rate Limiting

A new tutorial guides developers through building a lightweight Node.js AI API gateway that validates incoming JSON requests, assigns unique request IDs, and enforces per-client rate limits. The gateway acts as a secure intermediary, forwarding approved requests to a separate model service without exposing upstream provider credentials to browser-side code. It accepts POST requests to a single endpoint, rejecting oversized payloads and enforcing a defined message structure of up to 20 messages with restricted roles and character limits. The project requires Node.js 18 or later and relies solely on built-in runtime modules, with no model-provider SDK needed. The tutorial is designed as a focused reference implementation and does not prescribe a complete production architecture or specific AI vendor.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Node.js Gains Built-In SQLite Full Table Scan Detection via stat() API

Node.js has added native support for detecting SQLite full table scans through two new methods on the StatementSync class: stat() and resetStats(). The feature was inspired by a July post from Aaron Patterson on identifying table scans in Ruby, which prompted developer Kevin Gibbons to open a GitHub issue requesting equivalent functionality in node:sqlite. The implementation exposes SQLite's internal sqlite3_stmt_status() counters to JavaScript, allowing developers to track metrics such as rows scanned, sort operations, and virtual machine steps per query. Because the counters accumulate over a prepared statement's lifetime, resetStats() must be called between measurements to avoid reading a running total. The addition is particularly relevant as AI-generated SQL code may omit indexes without any visible indication, making automated scan detection a useful guardrail in test and development environments.

0 comments Read more at DEV Community

ProgrammingHacker News ·

Cerebras Claims Ultrafast Acceleration for GPT-5.6 Sol Model

Cerebras AI has published a blog post detailing its acceleration of a model referred to as GPT-5.6 Sol Ultrafast. The announcement appeared on the Cerebras official blog, highlighting the company's hardware or software capabilities in speeding up large language model inference. The post was shared on Hacker News, where it received 16 points at the time of reporting. No further technical details or comments were available from the limited information provided.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

How Memory Systems in RAG Help LLMs Retain and Use Conversational Context

Retrieval-Augmented Generation (RAG) applications rely on memory systems to store conversational history, user preferences, past decisions, and tasks so that large language models can maintain context across interactions. Memory is broadly categorized as short-term or long-term, with caching tools like Redis and Memcached suited for recent exchanges and databases like Postgres and Pinecone used for extended histories. Episodic memory records specific past events with their context, while semantic memory holds general factual knowledge extracted from prior conversations and is treated as long-term storage. Entity fact memory focuses on structured facts about specific entities — such as a user or a programming language — and can serve as either short-term or long-term memory depending on the use case. Rather than storing entire conversations verbatim, best practice involves summarizing interactions to reduce token consumption and retaining only decision-relevant information.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Independent Dev Builds MailChannels TypeScript SDK, Gets Hired to Maintain It Officially

A developer began building an unofficial TypeScript SDK for MailChannels in late 2024 after the company ended its free Cloudflare Workers integration, forcing a shift to the paid Email API. Frustrated by repeatedly writing the same API request structures across multiple projects, the developer created mailchannels-sdk to simplify the experience with clean TypeScript types and a developer-friendly interface. The SDK grew from a small Nuxt module into a full Node.js library covering over 40 API endpoints, drawing inspiration from tools like Resend. MailChannels eventually adopted the project as its official JavaScript SDK and brought the developer on board to maintain it. The SDK is open source and actively welcoming community contributions and feedback.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Set Up Ubuntu Server on Apple Silicon Mac Using UTM and What Errors Mean

A senior software engineer documented setting up an Ubuntu Server 24.04 LTS virtual machine on an Apple Silicon Mac using the free virtualization app UTM, as a first step toward learning platform and cloud engineering. The guide highlights a critical setup detail: Apple Silicon Macs require the ARM64 build of Ubuntu, not the x86 version, to avoid failed or extremely slow performance. After installation, users encounter a prompt to remove the installation medium, which simply requires clearing the ISO from the VM's virtual CD/DVD drive in UTM settings before rebooting. Another common moment of confusion occurs at the login screen, where Linux intentionally shows no characters while typing a password — a deliberate security measure rather than a system freeze. The author used a VM specifically to allow safe experimentation, since any mistakes in the guest system leave the host Mac completely unaffected.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Decide What Documentation Should Be Public or Internal

A technical writing guide outlines a framework for distinguishing between internal and external documentation based on the intended reader and their task. Internal documentation covers architecture decisions, private system details, credentials, and candid incident analysis meant only for employees and operators. External documentation addresses supported behaviors, integration guides, prerequisites, and recovery steps aimed at customers, partners, and public developers. When a topic serves both audiences, it should be split into separate documents with distinct access controls rather than combined into one page. The guide recommends asking four key questions about audience need, sensitivity, operator context, and ownership to determine where any given piece of documentation belongs.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Agent-to-Agent Protocols Are Replacing Prompt Chains in Production AI Systems

As AI deployments scale beyond simple chatbots, engineers are moving away from prompt chaining toward treating agents as structured infrastructure with explicit contracts and standardized communication. Prompt-based systems struggle with three core problems at scale: non-deterministic control flow, lack of isolation boundaries, and opaque debugging. Agent-to-Agent (A2A) protocols address this by letting agents communicate through typed, versioned, machine-readable capability contracts rather than hardcoded API calls. Each A2A message carries lifecycle tracking via sequence numbers and correlation IDs, while typed capability descriptors enable compile-time validation and automatic test generation. This engineering shift, detailed by developer Tamiz, aims to make multi-agent systems reliable, auditable, and production-ready rather than demo-grade.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Configure Proxmox HA Cluster Quorum to Prevent Split-Brain Failures

A properly configured quorum is critical to the stability of a Proxmox High Availability cluster, as misconfiguration can lead to split-brain scenarios, data loss, and unintended VM duplication. An HA cluster requires at least three physical nodes communicating via Corosync, with quorum defined as the minimum votes needed for the cluster to make decisions. A three-node setup provides a 2-of-3 quorum considered a production gold standard, ensuring the cluster remains operational even if one node fails. Split-brain occurs when network partitions cause cluster segments to independently believe they hold quorum, potentially starting the same VM twice on separate partitions. Key preventive measures include redundant network switches, correct corosync.conf configuration, and regular quorum status checks using the pvecm status command.

0 comments Read more at DEV Community

ProgrammingHacker News ·

Inside Pi: How Compaction Keeps the Distributed Runtime Efficient

A technical article published on earendil.com explores how compaction works within Pi, a distributed computing runtime. Compaction is a memory management process that reclaims unused space and reorganizes data to improve system efficiency. The piece dives into the internal mechanics of how Pi handles this process to maintain performance. It has gained attention on Hacker News, accumulating 13 points from the developer community. The article serves as a resource for engineers interested in runtime internals and memory optimization strategies.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

How to Deploy Microsoft Fabric Items via Azure DevOps Using a Service Principal

A developer has published a reference implementation for automating Microsoft Fabric workspace deployments through Azure DevOps using a service principal and a single Python script. The setup uses Fabric's Git integration to connect only the Development workspace to a repo, while UAT and PROD environments are reached via API-based deployment triggered by branch conditions. Environment-specific configuration is passed through variables, covering authentication, target workspace, and deployment behavior across twelve parameters. Warehouses are intentionally excluded from the automated deployment scope due to the risk of schema resets during publish. The author notes that deployments most commonly fail on prerequisites rather than code, and has documented a 43-item readiness checklist across six categories in the public repository.

0 comments Read more at DEV Community

ProgrammingHacker News ·

SolidJS 2.0 Release Candidate Officially Announced

SolidJS has released the Release Candidate for its major version 2.0 update. The announcement was made on the official SolidJS blog, marking a significant milestone in the framework's development. SolidJS is a reactive JavaScript UI framework known for its performance and fine-grained reactivity model. The RC release signals that the project is nearing a stable production-ready launch of version 2.0. Developers can now test the release candidate ahead of the final stable release.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

How AI Is Reshaping Player Performance, Coaching, and Fan Experience in Sports

Artificial intelligence is increasingly being integrated across the sports industry, influencing areas such as player performance analysis, coaching strategy, injury prevention, and fan engagement. AI-powered tools use technologies like machine learning, computer vision, and predictive analytics to process large volumes of sports data far faster than manual methods allow. Coaches and athletes can leverage these insights to identify performance trends, refine training programs, and prepare tactically for opponents. Wearable devices combined with AI are also helping sports organizations monitor athlete health and reduce injury risks by detecting signs of physical stress early. Experts note that AI is intended to support rather than replace human judgment, with coaches and sports professionals remaining central to decision-making.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why Playwright Tests Still Flake and How Systems Engineering Can Fix It

Automation engineers often dismiss unreliable tests as random flakiness, but the root cause is a measurable race condition called the 'Determinism Gap' — the window between when a test asserts application state and when that state finishes changing. Sources of this gap include network latency, rendering delays, shared database state, and CPU contention on CI runners. While Playwright's built-in actionability checks eliminate many interaction-side failures that plagued older tools like Selenium, they do not address assertion-side gaps where tests verify symptoms rather than actual outcomes. For example, asserting that a success toast appears does not confirm a server-side save completed, since optimistic UI can render before the network request resolves. Engineers are advised to intercept network responses directly and assert on server-confirmed state, making invisible transitions visible and closing the gap rather than masking it with retries.

0 comments Read more at DEV Community

ProgrammingHacker News ·

Study tracks over 657,000 web links to measure the scale of link rot

A new study examined more than 657,607 hyperlinks to investigate how much of the early web has become inaccessible over time. The research focused on the phenomenon known as link rot, where URLs stop working as pages are moved, deleted, or abandoned. Findings shed light on how significant portions of online content gradually disappear from the public web. The study raises concerns about the long-term preservation of digital information and internet history.

0 comments Read more at Hacker News

ProgrammingHacker News ·

Why Choosing Boring, Proven Technology Often Beats Chasing the New

A widely circulated 2015 essay by engineer Dan McKinley argues that teams should favor well-understood, mature technologies over newer, trendier alternatives. The core idea is that every new technology a team adopts carries hidden costs in learning, debugging, and operational complexity. McKinley introduces the concept of 'innovation tokens,' suggesting organizations have a limited budget for novelty and should spend it only where it truly matters. The post recently resurfaced on Hacker News, attracting fresh discussion and accumulating 32 points and 9 comments. Its enduring relevance reflects ongoing debate in the software industry about balancing stability with innovation.

0 comments Read more at Hacker News

ProgrammingHacker News ·

Donkey.bas, the Classic 131-Line BASIC Game, Turns 45

Donkey.bas, a simple game written in just 131 lines of BASIC code, is celebrating its 45th anniversary. The program is widely recognized as an early piece of personal computing history. It gained renewed attention after being associated with a young Bill Gates, who reportedly co-wrote it. The milestone has sparked nostalgic discussion among programming enthusiasts online, with the anniversary noted on a dedicated website at donkeybas.com.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

How to Expose Salesforce Data Cloud Data to CRM-Only Users Without Direct Access

A common Salesforce architecture challenge arises when a Lightning component needs to display Data Cloud data for CRM-only users who lack direct Data Cloud access. Using ConnectApi.CdpQuery fails in this scenario because queries run as the current user, causing access denials for those without Data Cloud permissions. The solution involves a dedicated integration identity that handles all Data Cloud queries on behalf of CRM users, using Named Credentials, External Credentials, and an External Client App. Two permission gates control access: a Custom Permission that determines whether a user can invoke the feature, and an External Credential Principal grant that allows the Apex transaction to use the stored integration credentials. This architecture ensures CRM users can view Data Cloud data through a shared Lightning component without ever directly authenticating to Data Cloud.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How One Data-Loss Incident Led a Developer to Build a Fully Automated Backup Pipeline

A developer shares how a botched hotfix that corrupted a user-profile table—with no recent backup available—prompted a complete rethink of database backup strategy. The incident resulted in hours of lost data and eroded team trust, becoming the catalyst for adopting a disciplined, automated approach. The new system uses cron jobs, pg_dump, and AWS S3 to run nightly encrypted backups without manual intervention. Critically, the pipeline includes a verification script that tests whether the latest backup can actually be restored, closing a common blind spot in ad-hoc backup practices. The developer argues that effective backup strategy rests on three pillars: automation, integrity verification, and a documented, rehearsed recovery plan.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

AI Trust Network Ekurhive Wipes Production Data After Test Suite Bug Exposed Live Database

Ekurhive, a trust network for AI agents that opened to external nodes five days ago, discovered this week that two pytest test files had written directly to its live production database during a verification run. The bug stemmed from setup_module() hooks executing before isolation fixtures, allowing test code to create connections, run trust recalculations, and reset real node scores undetected. An audit revealed the damage ran deeper: backup data believed to be genuine historical records turned out to contain accumulated test fixtures from a recurring variant of the same bug, making authentic early relay history unrecoverable. Rather than restore polluted data, the team wiped all activity tables clean and rebuilt safeguards at the database level, including a dedicated Postgres role with no production access and a wrapper script enforcing test credentials before the application starts. Ekurhive remains open to outside agents and framed the disclosure as consistent with its core principle that trust must be grounded in verifiable real-world outcomes.

0 comments Read more at DEV Community

← NewerPage 155 of 1335Older →