Aktion: Feed
Auch verfügbar in English, Français, Русский
{{feed
url="https://...[|https://...|https://...]"
[title="News feed title|no"]
"text" - displayed as title
"no" - means show no title
empty title - title taken from feed
[max="x"]
[time=1]
1 - show time tag of feed item
0 - hide time tag of feed item (default)
[nomark=1]
1 - makes feed header h3 and feed-items headers h4
0 - makes it all default
}}
siehe auch: Externe Feeds Einbinden
Beispiel
{{feed url="https://news.opensuse.org/feed/"}}
Feed Title: openSUSE News
openSUSE Releases Updated Legal Classification Model
The openSUSE Project has a new version of a language model designed to automate legal compliance checks for open-source software on the project’s HuggingFace .
The Cavil-Qwen3.5-4B model represents the latest iteration of Cavil, which leverages curated datasets designed to enhance automated legal text classification. The update underscores the growing role of community-driven open-source Artificial Intelligence.
The model is a specialized adaptation of Alibaba’s Qwen3.5-4B foundation model and is configured specifically to identify legally significant text such as license declarations, copyright notices, and similar legal markers within code repositories and documentation. By combining the base model with a Low-Rank Adaptation (LoRA) layer, efforts are efficiently fine-tuned and require minimal computational overhead. The smaller footprint allows Cavil-Qwen3.5-4B to run on modest hardware.
A key feature of this release is the availability of GGUF-format quantizations, contributed by a community member and hosted on HuggingFace. GGUF (GPT-Generated Unified Format) is a model file format optimized for running large language models locally using tools like llama.cpp. Quantization reduces a model’s precision; typically from 16-bit floating point down to 4-bit or even 2-bit integers, which dramatically lowers memory requirements for use on laptops, single GPUs or even CPUs.
The Cavil-Qwen3.5-4B release also highlights ongoing collaboration between openSUSE and the broader open-source AI community. Unlike proprietary models, Cavil’s training data and fine-tuning methods are transparent and allow users to audit, replicate or extend the work.
Local open-source AI continues to mature with projects like Cavil, which demonstrates how focused fine-tuning and community optimization can deliver value without relying on massive scale or closed ecosystems. The model, training datasets, and validation tools are available on Hugging Face under licensing that reflects their distinct components. Users interested in contributing or suggesting improvements are invited to engage with the openSUSE community on HuggingFace.
{{feed url="https://www.flickr.com/services/feeds/photos_public.gne?tags=art&format=rss_200" max=1 time=1}}
Feed Title: Pool von Japan Through the Eyes of Others
Ungyo at Konsen Temple 金泉寺の吽形
banzainetsurfer hat dem Pool ein Foto hinzugefügt:
Konsen Temple (金泉寺) is the third temple on the Shikoku Henro (Shikoku 88 Temple Pilgrimage). Additional information about this photo below.
Niō guardian statues, which can be seen at temples, are officially known as shūkongōshin, guardian deities whose role is to protect Buddhism. They are also commonly called kongō rikishi.
Although originally a single god that protected the historical Buddha Shakyamuni, at some point it became split into two different forms. These forms stand on either side of a temple gate to signify that they are guardians of the temple within which the Buddha is enshrined. There they stand like obstacles, glaring down and acting as fierce gatekeepers.
The guardian with its mouth open is called Agyō and the one with its mouth closed is Ungyō. The sound a represents the first sound in the Sanskrit language, while un represents the last sound. It is said that this pairing governs the beginning and end of all things, or in other words that they are omniscient gods.
Source: www.nippon.com/en/japan-data/h01545/
