Developer builds free open-source app to preserve GitHub traffic stats beyond 14-day limit

·2 views

A developer frustrated by GitHub's 14-day traffic data retention limit has created Repo-rter, a free and open-source desktop application for Windows, macOS, and Linux. The app fetches GitHub repository traffic data and caches it locally, allowing maintainers to retain unlimited historical stats. Built with Tauri and Next.js, the lightweight application (~15MB) also tracks release asset downloads and supports exporting reports in Markdown format. Users can set it up in about 30 seconds by connecting a read-only GitHub personal access token. Version 0.3.0 is currently available on the project's GitHub releases page.

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)

Developer Builds Fighting Game Where AI Models Battle Based on Real Architecture

A developer on DEV Community has created 'Model Kombat,' a retro-style 2D fighting game in which popular large language models such as GPT-4o, Claude Opus, DeepSeek, and Mixtral compete as playable characters. Each fighter's visual complexity and abilities directly reflect real-world AI engineering concepts, including parameter count, context window size, and reasoning token generation. Game mechanics like the Ki bar simulate time-to-first-token latency, while a 'Context Eviction' penalty represents what happens when a model's context window overflows. Mixture-of-Experts models dynamically activate specialist nodes during combat, granting buffs tied to text, math, or vision capabilities. The project aims to make abstract AI benchmarking concepts more intuitive and engaging through interactive gameplay.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

GSoC 2026 Week 5: Developer Completes CircuitVerse Import Pipeline After Illness

A Google Summer of Code 2026 contributor working with CircuitVerse completed Week 5 of their project between June 22 and June 28, returning to full productivity after missing the previous week due to illness. The major milestone this week was finishing the implementation of importCanonical.ts, a file responsible for reconstructing circuits from canonical JSON data back onto the user's canvas. The import pipeline supports multiple circuits and uses Kahn's Algorithm to resolve subcircuit dependencies in the correct loading order, mirroring the existing export pipeline. JSON schema validation has been intentionally left out for now and will be added once the canonical format is finalized. Three foundational pull requests remain under mentor review on GitHub, and the new import code will be submitted as a separate PR only after those core changes are merged into the main branch.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How One Developer Scaled a Static Site to 4,400 Pages While Staying SEO-Friendly

A developer built Luxury Hotel Offers, a fully static website generating 4,400 HTML pages at build time from over 3,400 hotel listings, with no server-side rendering or runtime database. To avoid Googlebot's 2 MB page limit, the site caps initial HTML at 400 cards and loads remaining listings via separate fragment files. A content-hashing system ensures only genuinely updated pages receive new lastmod timestamps and IndexNow notifications, preventing search engines from flagging bulk updates as spam. Mobile filtering performance was improved through pre-indexed search attributes, batched DOM updates, and an Edge Runtime API for complex queries. Images sourced from Notion's expiring S3 URLs are downloaded, optimized with Sharp, and stored persistently in Cloudflare R2 to avoid redundant processing across builds.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Cyrillic look-alike characters can silently bypass AI prompt safety filters

A technique called homoglyph evasion allows attackers to bypass AI content filters by substituting standard Latin characters with visually identical Cyrillic or Greek Unicode equivalents. Because keyword and regex filters match raw bytes, a jailbreak phrase written with look-alike characters passes through undetected even though it appears identical to a human reader. Security researchers recommend normalizing input text into a canonical ASCII form before running detection rules, without altering the original text sent to the model. This normalization process involves stripping zero-width characters, applying NFKC Unicode normalization, and mapping known homoglyphs to their Latin counterparts. An open-source, OpenAI-compatible security gateway implementing this approach alongside roughly 170 other detection rules has been released to help developers address the vulnerability.

0 comments Read more at DEV Community