Mobilerun Uses AI Agents to Automate Mobile Apps That Lack Public APIs

·1 views

Most consumer apps in banking, social media, and messaging do not offer public APIs, making automation difficult for developers. Traditional UI testing tools like Appium and Selenium proved too brittle, breaking whenever an app redesigned its interface. A new open-source framework called Mobilerun addresses this by combining LLM vision models with the device's accessibility tree to let an AI agent read screens and execute multi-step workflows. The framework supports both Android and iOS, works on real physical devices to bypass emulator-detection used by high-security apps, and offers a CLI, Python SDK, and hosted cloud option. By treating mobile automation as an intent-driven task rather than a structural one, Mobilerun aims to fill the gap left by missing or incomplete APIs.

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)

How to Build a Production-Grade Apache Reverse Proxy with Load Balancing and Failover

Apache is widely used as a TLS-terminating reverse proxy in front of backend servers like Node.js, Gunicorn, or Tomcat, but basic ProxyPass configurations often cause 502 errors and crashes under real traffic load. A critical security requirement is setting ProxyRequests Off to prevent Apache from acting as an open relay that attackers can exploit for DDoS or spam. Connection pooling through proxy directives helps Apache recycle TCP connections during traffic spikes, avoiding backend connection exhaustion. Administrators must also distinguish between the Timeout directive, which governs client-to-Apache communication, and ProxyTimeout, which controls how long Apache waits for a backend response. Finally, headers such as X-Forwarded-Proto and ProxyPreserveHost must be configured so backend applications correctly identify the original client IP and protocol, ensuring proper rate limiting and routing.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How ChatGPT, Perplexity, and Google AI Overviews select sources to cite

AI answer engines like ChatGPT, Perplexity, and Google AI Overviews use retrieval-augmented generation pipelines that pull candidate pages, rerank them, and cite only a small fraction in the final response. An AirOps analysis of roughly 15,000 prompts found ChatGPT cites just 15 percent of retrieved pages, meaning most content is evaluated and silently discarded. ChatGPT relies heavily on Bing's index, with one study finding over 87 percent of its citations matching Bing's top organic results for the same query. Perplexity uses hybrid keyword and vector retrieval combined with multi-stage reranking that weighs freshness, domain authority, and topical fit, favouring pages on established trusted domains. Google AI Overviews differ by citing specific extractable passages rather than whole pages, with roughly 96 percent of cited sources carrying strong expertise, authoritativeness, and trustworthiness signals.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How a Static Open Source Website Evolved into a Live Intelligence Platform

The team behind WebiU 2.0 set out to redesign a static website for an open source organization, only to discover the real challenge was a data management problem, not a frontend one. As repositories multiplied and contributor activity changed daily, manually keeping the site up to date became unscalable. The developers shifted their approach by treating GitHub as a live data source rather than a static reference, enabling dynamic collection of repository metadata, contributor stats, and project health indicators. This architectural rethink transformed the platform from a simple informational site into a system capable of reflecting real-time open source activity. The experience highlights how growing open source organizations eventually outgrow static web infrastructure and require purpose-built data pipelines to stay accurate.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Keenmate Web DateRangePicker v2.0.0-rc02 Adds New Lock/Unlock Command

Keenmate has released v2.0.0-rc02 of its web-daterangepicker component on npm. The update introduces a new lock/unlock command designed to support booking-style workflows. With this feature, users can select dates and preview an estimated price before confirming, after which the picker becomes locked pending an external update. The lock can be applied to the entire component or just a portion of it, depending on the parameter passed. The release candidate is currently available on the npm registry for developers to test.

0 comments Read more at DEV Community