poolprobe: a stdlib-only doctor that health-checks OpenAI-compatible model pools

·9 views

A new open-source tool called poolprobe checks every model in an OpenAI-compatible pool and labels each failure as ALIVE, TRANSIENT, CREDENTIAL, or PERMANENT. Its design prevents removal of working models, since a model is only declared dead after five independent checks agree. The project is built on the Python standard library alone and is installable via pip under the MIT license. According to the author, the tool only reads the data it is pointed at and sends nothing externally. The author also cautions that its rules can yield false positives, so it is meant to flag candidates for review rather than serve as a definitive judgment.

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)

Free FMCSA census query surfaces newly registered trucking companies

A developer shared a method for pulling newly registered trucking companies from the FMCSA carrier census, which is published as open data on data.transportation.gov. The approach uses a SoQL query over plain HTTP, filtering active records by registration date and state, and requires no API key for light use. The author notes that the raw dataset lacks built-in deduplication, operating authority status, and normalized fields, which must be handled separately. To address these gaps, the author built a paid Apify actor that joins authority data, applies filters, avoids repeat deliveries, and exports results in JSON, CSV, or Excel. The article notes the author's bias, since they built the tool they promote.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Expo Router navigation slows past 40 nested routes, with fixes for layout re-renders and focus

A team migrating a multi-tenant dashboard from React Navigation to Expo Router found a performance drop at around 42 nested routes, with animations falling from 60fps to 38fps and memory rising 140MB on Android mid-range devices. Benchmarks on a Pixel 5 and iPhone 12 using Expo SDK 52 and React Native 0.76 showed navigation time and frame drops rising sharply as route depth increased. The team attributed the problem to re-renders of the layout tree on navigation events, compounded by runtime TailwindCSS class recalculation in nested layouts. Replacing dynamic class props with static StyleSheet references cut navigation time at 50 routes from 187ms to 94ms and reduced frame drops per transition from 9 to 3. The team also found that modal routes overlaying tab navigators broke screen reader focus, which they addressed by wrapping the modal in a container with accessibilityViewIsModal.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Plex Media Server flaw CVE-2026-96655 can be checked in logs for prior exploitation

Patching Plex Media Server to 1.43.3.10861 closes the SSRF in CVE-2026-96655, but it does not reveal whether the flaw was already abused. The bug allows an authenticated user to make the server fetch an attacker-chosen target via the universal transcoder endpoint. Investigators are advised to compare endpoint activity against a baseline of normal household use and to look for internal network connections from the Plex host that it would not normally make. Failed transcode sessions can serve as a weak triage signal but are not conclusive evidence. If compromise is suspected, logs and connection records should be preserved, the involved account disabled, and any exposed credentials rotated before patching.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Video generation models still face a wide gap between demos and production use

Video models can produce impressive short clips in marketing demos, but production pipelines run into latency problems, weak frame-to-frame consistency, limited camera control, and mandatory brand-safety review. Treating video generation as simply a longer form of image generation tends to underestimate costs and iteration cycles. Teams that successfully ship multimodal features usually begin with narrower tasks such as understanding clips, captioning meetings, or extracting frames for search. Assisted editing, storyboards, and animatics are also more practical than generating final pixels for regulated brands. The article notes that legal and brand review often takes more time than GPU inference, and it recommends tools with controllable parameters and evaluation methods for temporal artifacts.

0 comments Read more at DEV Community