Developer Rebuilds CI/CD Platform OdinRun With Self-Healing and AI Features

·2 views

A developer has documented the engineering journey of rebuilding PipelineOS into OdinRun, a self-hosted CI/CD runtime designed for reliability, observability, and auto-remediation. Early versions suffered from Docker stream corruption, zombie containers, and flaky runner registration, prompting a major refactor that split the monolithic runner into focused, single-responsibility components. Subsequent milestones introduced an in-process domain event architecture to replace chatty HTTP calls, and pluggable storage abstractions supporting both SQLite and MongoDB backends. Real-time telemetry was added to stream CPU, memory, and log data live to the UI via Server-Sent Events and WebSockets. The latest phase introduces rule-based auto-remediation, with AI assistance positioned as a secondary layer after deterministic logic.

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)

pictostl Turns Photos Into 3D-Printable Meshes via Browser Tool

pictostl is a browser-based tool that converts JPG, PNG, or WebP images into 3D meshes suitable for printing and modeling workflows. Users can upload images up to 10 MB, optionally add multiple views, and select from Basic, Pro, or Ultra quality tiers. The tool exports meshes in STL, GLB, OBJ, and PLY formats, with an interactive preview and adjustable dimensions before download. New verified accounts receive two free credits, covering one Basic generation and STL export, while additional generations require purchased credits. Since unseen surfaces are algorithmically inferred, the developers recommend reviewing all generated meshes carefully before sending them to a printer.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Silent Cron Job Failure Left Production Monitor Dead for Five Days Undetected

A developer running WhatsApp automation for businesses discovered that a watchdog cron job, designed to alert when messaging sessions dropped silently, had never executed once despite appearing correctly configured. The crontab stored the job as expected, and a separate log file showed recent successful entries — but those entries turned out to be from manual test runs on installation day, not scheduled executions. The root cause was a sudo permission or output-redirect issue causing the cron job to emit errors on stderr every ten minutes, which cron attempted to email but discarded silently due to no mail server being installed. A key diagnostic clue was the syslog message 'No MTA installed, discarding output' appearing alongside every CMD entry, indicating the job was running but failing to capture output as intended. The incident highlights that a populated log file and a visible crontab entry do not confirm a scheduled job is actually executing successfully.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Union Alpha AI Model Revealed as Pareto by unbiased.ai, Free Access Ends Early

A mysterious AI model called Union Alpha, released anonymously on OpenRouter on September 16, 2026, was identified the following day as Pareto, a blended AI model developed by unbiased.ai. The model was initially offered free for approximately one week, but the free tier was shut down ahead of schedule on September 17 after demand surged to one billion tokens per minute within a single day. AWS tripled compute capacity overnight but still could not keep up with the load, prompting the company to switch to paid access. Pareto is described as a blended model that runs multiple AI models in parallel for every request and synthesizes a single response, distinguishing it from a standard model router. The company's official launch is planned for October 10, 2026, and current pricing for the latest version, Pareto 26.9, stands at $2.50 per million input tokens and $7.50 per million output tokens.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

BCA Student Builds Python CLI Contact Book App with Local JSON Storage

A second-year BCA student has developed a command-line contact book application built entirely in Python as their fourth personal project. The app supports full CRUD operations — adding, viewing, searching, and deleting contacts — through an interactive menu interface. All contact data is stored locally in a JSON file, ensuring persistence between sessions, while input validation enforces 10-digit phone numbers and proper email formats. The project also incorporates try-except error handling to prevent crashes from bad or missing inputs. The student has published the full codebase on GitHub and is seeking community feedback on potential next steps, such as adding an edit feature or migrating to a SQLite database.

0 comments Read more at DEV Community