REST API news and engineering summaries

Bose has officially released the API documentation for its SoundTouch speaker line, offering a lifeline to users facing the impending end-of-life (EoL) status of these premium devices. Originally, Bose announced that these systems would lose cloud connectivity and app support by February 2025, effectively bricking several multi-room features. However, the company has clarified that Spotify Connect and AirPlay 2 will remain functional, allowing continued wireless streaming. Furthermore, a local version of the SoundTouch app will be released in 2026 to facilitate control without cloud dependency. This move addresses significant customer backlash regarding the planned obsolescence of expensive hardware, ensuring that these high-end speakers retain functionality through local integration and open documentation even after official manufacturer support ceases.

RFC 10008 introduces the HTTP QUERY method, providing a safe and idempotent way to perform queries that are too complex for URI encoding. Unlike POST, QUERY allows for automated retries and caching. It supports content negotiation and URI assignment for query results, bridging the gap between GET and POST while maintaining resource safety.

The author successfully built a desktop environment to boot a Tesla Model 3 Media Control Unit and touchscreen. By sourcing salvage parts, navigating internal wiring schematics, and replacing a damaged power controller chip, they created a setup to explore the vehicle's network interfaces, REST API, and firmware, all in preparation for the Tesla bug bounty program.

LocalSend is an open-source, cross-platform application for secure, local file and message sharing. It uses HTTPS encryption and a REST API to enable direct device-to-device communication without an internet connection or external servers. The project supports various operating systems and provides detailed guidance on installation, network configuration, and open-source contribution.

Cloudflare launched Web Search API in beta, enabling AI agents and applications to ground responses in live Internet information. It initially supports Ceramic.ai, Exa, and Linkup, each offering Zero Data Retention through Cloudflare and following verified bot crawling standards. Running through AI Gateway, requests appear in gateway logs and use provider list pricing without markup, with support for REST API, Workers AI bindings, and customer-owned provider keys.

Developers are abandoning popular API tools like Postman and Insomnia due to corporate bloat, forced cloud-sync, and paywalled collaboration features. Users are shifting toward local-first, Git-native alternatives like Bruno that prioritize performance, offline capabilities, and developer privacy over enterprise-focused business models.

The Decisions API is in public beta and evaluates text, images, or both about 10x faster than the Responses API, returning typed answers for conditions, fixed choices, or rubric scores. It uses the POST /v1/decisions endpoint and currently supports only gpt-6-luna. Its predicate, choice, and score questions support classification, routing, prioritization, and multimodal checks; responses include probabilities and, for choice and score, confidence. Input costs $0.10 per 1M tokens, with no output-token charges.

http.cat is a popular and creative web service that provides a unique way to visualize HTTP status codes using high-quality cat images. Developed by Rogerio Vicente with original artwork by Tomomi Imura, the platform allows developers and enthusiasts to access visual representations of almost every HTTP response code, ranging from the standard 200 OK to more obscure errors like 418 I'm a teapot and 451 Unavailable For Legal Reasons. Users can access these images directly via a Rest-API style URL structure, making it a favorite tool for developers to add some humor to documentation, error pages, or Slack notifications while maintaining technical accuracy regarding the HTTP protocol.

A benchmark comparing vision-based AI agents to structured API agents reveals that vision agents are approximately 45 times more expensive and significantly slower. Using Claude Sonnet, vision agents required 550k tokens and 17 minutes to complete tasks, while API agents used only 12k tokens and 20 seconds, demonstrating that structural interface design drastically impacts agent efficiency and cost.

The Art Institute of Chicago's API includes a boolean field called has_not_been_viewed_much, identifying artworks viewed fewer than 200 times on their website since 2010. This metric highlights overlooked pieces in their collection, encouraging users to discover and explore lesser-known art from their digital archives.

The March 2026 updates for Postman signify a radical shift from its origins as a focused API testing client toward an all-encompassing platform. Key additions such as AI Agent Builders, managed MCP Servers, and integrated terminals aim to transform Postman into a centralized hub for AI-powered API development. However, these changes raise significant concerns regarding feature bloat, performance degradation, and UI complexity. The transition toward a cloud-first model suggests a strategy of vendor lock-in, potentially stifling innovation by moving core business logic into a proprietary ecosystem. To maintain architectural independence and control, developers are encouraged to weigh the risks of dependency on third-party infrastructure against the simplicity of lightweight, open-source alternatives for API verification.

Effective APIs prioritize long-term stability over initial elegance. By maintaining strict boundaries, avoiding tight coupling to frontend structures, and minimizing unnecessary data exposure, developers can prevent accidental dependencies. Building 'boring' but honest APIs that account for evolving requirements reduces technical debt and maintenance challenges, ultimately fostering trust and reliability across team integrations.

GitHub built an aggregate star-history endpoint that returns up to 30 weeks of daily counts without stargazer identities. It filters stars by an indexed time range before grouping, reducing `react/react` history retrieval from about 2,511 requests and 6 million repeated row scans to about 24 requests examining roughly 251,000 rows. The design addresses scraping risks and database load, though unusually dense weeks can still be expensive.

GraphQL can consolidate requests and let clients select only needed fields, but it adds data-layer coupling, validation complexity, security risks, and aggregate-query limitations. It is most valuable for large platforms serving diverse clients. Simpler APIs can achieve similar results through SQL/NoSQL joins, optional fields, API flags, and HTTP/2 or HTTP/3 multiplexing, often with better performance.

This content provides documentation and technical details for REST API and gRPC services, including information on pricing and a search interface to facilitate navigation and developer workflows.

To make a Blink doorbell audible at home, the author built an elaborate automation chain: Alexa triggers a virtual SmartThings light, SmartThings sends webhooks to a VPS, and Home Assistant polls it to play audio on Google Home. The setup avoids subscriptions and inbound access, but SmartThings’ new $4.99 monthly API fee threatens its future.

Bluesky launches Bluesky Protocol Services, a unified developer portal for its AT Protocol infrastructure, and Jetstream v2 with Network Replay, enabling filtered historical snapshots and seamless live-stream catch-up. It also releases TypeScript and Go SDKs, rebases the Bluesky TypeScript SDK on lexicons, updates API references, and keeps services open-source and self-hostable.

RFC 10008 introduces the HTTP QUERY method to handle complex read-only requests. Unlike GET, which uses bulky query parameters, or POST, which lacks safe, idempotent semantics, QUERY allows request bodies for complicated queries. While promising, widespread adoption is limited, and GET remains preferred for shareable links and simple filtering.

Janus is a single Go binary that runs local .gguf LLMs through llama.cpp using Vulkan on AMD, Intel, or NVIDIA GPUs, with CPU fallback, and exposes an OpenAI-compatible API. It solves the need for a lightweight local server without Python, Docker, or Ollama. Its distinguishing features include hot-swapping models, automatic GGUF chat-template detection, streamed chat completions, thinking-model reasoning output, and integration with Cursor, Cline, and other OpenAI clients.

GitLab.com will align rate limits with subscription tiers: Free and unauthenticated traffic change on October 19, 2026, while Premium and Ultimate change in January 2027. Authenticated requests receive plan-based limits; anonymous requests are capped at 60 per hour per IP. Users should authenticate, optimize API usage, and handle HTTP 429 responses. Self-Managed and GitLab Dedicated are unaffected.