WordPress 7.1 breaks ACF block editors by forcing iframe canvas for all posts

·1 views

A change in WordPress 7.1 removed the conditional logic that previously kept the block editor canvas un-iframed when legacy ACF blocks were present. Prior to 7.1, the editor skipped iframing if any block in the post used API v2 or lower, which effectively protected sites using default ACF blocks. In 7.1, that check is gone and the editor forces iframing unconditionally, causing ACF custom fields to shift to the sidebar or silently fail to save. Updating ACF does not resolve the issue, as the root cause is a core JavaScript change in WordPress itself. Developers can check which version they are on by running a grep for 'disableIframe' in wp-includes/js/dist/, where four results indicate 7.0.x and zero results confirm the 7.1 breaking behavior.

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 AI Shipment Agent Using Telnyx SMS, Voice, and Inference APIs

A developer has released an open-source Python and Flask application that turns a shipment into an AI-powered communications agent, rather than relying on static tracking pages. The app, called ShipmentAgent, uses Telnyx Messaging to send proactive SMS updates when carrier status changes and processes customer replies using Telnyx AI Inference for context-aware responses. Inbound phone calls to the Telnyx number are also handled by the same agent, ensuring SMS and voice interactions share identical shipment context. The project is available on GitHub under the Telnyx code examples repository and requires a Telnyx API key along with a configured messaging profile to run locally. The developer notes the sample is intentionally minimal and recommends additions such as persistent storage, customer authentication, and human escalation for production logistics use cases.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

moeinGTS Launches Open-Source AI Suite Spanning Edge LLMs and Vision Adapters

Developers Ali and Arshiya Sohrevardi have introduced the moeinGTS ecosystem, a family of open-source AI models designed for local, resource-efficient deployment. The lineup includes five specialized models ranging from a 1.5-billion-parameter lightweight LLM for edge devices to GTS-1, a flagship model built for complex reasoning and multi-agent coordination. A fine-tuned vision adapter called moeinGTS-kamal-1 handles image generation, while moeinGTS-paspan focuses on security, prompt injection defense, and output safety filtering. All models are optimized for low-VRAM environments using GGUF quantization formats and are compatible with frameworks such as Ollama, Hugging Face Diffusers, and PyTorch. The suite is publicly available on Hugging Face and Ollama Hub, and is designed so the individual models can work collaboratively within a single agentic pipeline.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Build Secure GitHub Actions Workflows with Least-Privilege Principles

GitHub Actions functions as a combined event system, workflow scheduler, compute environment, and credential broker, making security design critical for DevOps teams. The core security concern is limiting what authority untrusted inputs — such as pull request content or third-party action outputs — can exercise within a workflow. Experts recommend scoping permissions explicitly at the job level, with build jobs using read-only access and deployment jobs gaining elevated rights only after passing protected-environment checks. OpenID Connect is preferred over long-lived credentials, as it issues short-lived, claim-bound tokens at runtime, reducing exposure. Additional best practices include pinning third-party actions to immutable commit SHAs, separating build from deployment stages, and promoting a single verified artifact across environments rather than rebuilding for each.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

True ownership means surfacing problems, not just fixing what lands in your queue

A software engineering post on DEV Community argues that real ownership is not about grabbing every task or waiting for a ticket, but about ensuring a problem does not stay invisible once you have noticed it. The author identifies a common pattern where recurring issues appear in observability dashboards and team channels but never become formal cards, eventually escalating into incidents. To address this, they propose a four-step framework — See, Understand, Expose, Own — designed to move engineers from task-level thinking toward system-level awareness. The framework distinguishes ownership from implementation, meaning a developer can own a problem by documenting it, raising it with the right person, or explicitly accepting the risk as a conscious decision. The author cautions that if ownership becomes synonymous with coding the fix, senior engineers become bottlenecks and the model burns out talented people.

0 comments Read more at DEV Community