Made in PalermoTry Calogera AI →

Artificial intelligence, made in Sicily.

Calogera writes, translates and plans your day, even without internet. Simple tools, made for people.

Calogera-1.0 · 27B onlinealso offline on your deviceApache 2.0models and licences
27Bparameters in Calogera-1.0
500+books in its library
0internet needed, offline
In 20 seconds

Watch Calogera at work.

  • Talks the way you talkUnderstands everyday Italian and Sicilian.
  • Even without internetDownload it once to your device and it works offline.
  • With sourcesWhen it uses a book or a page, it shows it under the answer.
Products

The future speaks Sicilian too.

Clear, fast technology that understands everyday Italian and Sicily.

Calogera

online and offline

The first Sicilian AI, offline too. It writes, translates, summarises and plans your day; when it comes to Sicily, it knows what it’s talking about. It works even without internet.

calendar · mail · notebooks · weather
your data stays on your device
Open Calogera →

Coming soon

We’re building a new Trynakria tool. Coming here soon.

What you can do

Everyday things, made simpler.

Write and translateTexts, summaries and translations, or help with a problem.
Study with NotebooksAdd texts, files, websites or videos and ask questions about them.
Plan your dayEvents and tasks in the Calendar, ready to export.
Mail and messagesEmail and message drafts ready in seconds.
Understand documentsDeadlines, amounts and pagoPA code ready to copy.
Discover PalermoHistory, food and events in Sicily, with sources.
About us

Born in Palermo, for everyone.

Trynakria wants to bring technology and artificial intelligence within everyone’s reach, starting from our own land.

Arturo Cannizzaro

Founder and Lead Developer

Born in Palermo, he conceived and built Trynakria entirely on his own at the age of 16.

He personally leads web development, software architecture and the technical evolution of the platform, with the goal of creating innovative solutions for the region.

Roadmap

Where we are, where we’re going.

  • Done
    Trynakria website onlineThe first version of the platform is available.
  • Done
    Calogera, offline tooThe first Sicilian AI that works without internet.
  • In progress
    Calogera keeps growingNew features and continuous improvements.
FAQ

The questions you ask us most.

What is Calogera?

An artificial intelligence developed in Palermo by Trynakria. It writes, translates, summarises, helps you study and plans your day. When it comes to Sicily, it knows what it’s talking about. It works best in Italian.

Do I need an account?

No, you can use it right away at calogerai.com. With an account you find your chats on all your devices.

Does it really work without internet?

Yes. Calogera downloads once to your device (about 1.3 GB on a computer, about 530 MB on a phone) and then answers offline too. Offline it is a bit less accurate than the online version.

Does it understand Sicilian?

Yes, it understands everyday Sicilian and can answer in Sicilian. It’s still learning: every new version speaks it better.

Where do my conversations go?

Online, messages go to the Hugging Face servers where Calogera runs, only to generate the answer. Offline they stay on your device. Account holders can choose whether to donate conversations to improve Calogera, and can turn it off at any time. All the details are in Models and licences and in the privacy notice.

Can Calogera make mistakes?

Yes, like any artificial intelligence. For important things (health, money, law, deadlines) always check the sources or ask an expert.

Is there a phone app?

Coming soon to Google Play. Meanwhile, open calogerai.com on your phone and choose “Add to Home Screen”: it opens like an app.

Who is behind Trynakria?

Arturo Cannizzaro, from Palermo. He created Trynakria and Calogera on his own, at 16.

Contact

Let’s build it together.

Got an idea, a question or want to collaborate? Trynakria is looking for developers, designers and content writers. Every message is read by Arturo himself.

Ideas, partnerships and presscontatti@trynakria.com Help with Calogerasupporto@trynakria.com
← Back to home

Models and licences

The AI models Calogera uses, where they come from and under which licences. Today Calogera answers with Calogera-1.0 (online); without internet it uses Calogera-0.0 or Calogera Mini on your device.

Calogera-1.0 · online, the one you use nowThe model Calogera uses automatically when you’re connected. Developed by Trynakria (Palermo): a fine-tune (LoRA) of Qwen3.5-27B by Alibaba Cloud (Qwen Team), with 27 billion parameters, released under the Apache 2.0 licence; the base weights are those published by Unsloth. It runs on Hugging Face GPU servers (ZeroGPU): messages are sent there only to generate the answer and are not used by Hugging Face for anything else. Libraries: Transformers and PEFT (Hugging Face, Apache 2.0 licence). If the server doesn’t respond, Calogera switches to the on-device model by itself. Tap the model name in the chat to choose yourself. Calogera Lens also uses it to read and explain the documents you photograph: photos and text go to the Hugging Face server only for the answer, and your document archive stays in your browser.
Calogera-0.0 · on your device, no internetDeveloped by Trynakria (Palermo). A fine-tune of Qwen3.5-2B by Alibaba Cloud (Qwen Team), released under the Apache 2.0 licence. It runs directly in the browser: it downloads once (about 1.3 GB) from Hugging Face and then works offline too. It’s a small model: for important facts, check the sources. Engine: wllama (llama.cpp for the browser, MIT licence).
Calogera Mini · on phonesOn phones, where there isn’t enough memory for Calogera-0.0, it uses Qwen3.5-0.8B by Alibaba Cloud (Qwen Team, Apache 2.0 licence), in the GGUF version prepared by Unsloth, with Calogera’s instructions. It isn’t trained by Trynakria yet: it’s lighter (about 530 MB) but less accurate. It runs on the device and works offline too.
Calogera Vista · photos without internetWhen it works on your device, without internet, Calogera looks at photos with Calogera Vista: it describes what’s in the image and reads any written text, then uses that to answer. It downloads once (about 300 MB) and photos never leave your phone. When you’re online, Calogera-1.0 looks at photos directly, only to answer.
Calogera’s libraryTo answer, Calogera consults a library on its server that grows over time: all 504 Italian-language books from Project Gutenberg (public domain: Pitrè, Amari, De Roberto, Capuana, Verga, Pirandello, Manzoni, Dante, De Amicis, Garibaldi and many more), texts from Italian Wikisource (public-domain works, laws and historical documents) and thousands of articles from Italian Wikipedia (1 November 2023 version, CC BY-SA 4.0 licence, authors: Wikipedia contributors). When it uses a text from the library it cites it under the answer, in “Sources”, with a link to the original.
Calogera Voce · voice and audio filesTo understand what you say into the microphone and to transcribe audio files you attach (voice notes, recordings, lectures) Calogera uses Calogera Voce, which runs on your device: it downloads once (about 150 MB) and audio is never sent to any server. Calogera only receives the transcribed text, which may contain listening errors.
Photos in answersWhen it can, Calogera shows one or two photos from Wikimedia Commons, Wikipedia’s free image archive, under the answer. Each photo has its own licence (public domain, CC BY, CC BY-SA…) and author: the link under the photo leads to the image page, where they are listed. Calogera does not use random photos from the web.
  • Training data: examples written by Trynakria; Italian conversations from Aya Dataset (Cohere Labs) and OpenAssistant oasst2 (OpenAssistant / LAION), Apache 2.0; excerpts from Wikipedia, Wikivoyage, Wikibooks, Wikiquote and Wiktionary in Italian (CC BY-SA and GFDL); Sicilian sentences from Napizia (CC BY-SA 4.0); ITA-Base (Apache 2.0); human-written datasets machine-translated into Italian in the Aya Collection (Dolly, Flan, PIQA). The full list with licences is in each model’s NOTICE file.
  • User conversations: for account holders, “Help train Calogera” is on by default and can be turned off at any time from Account → Privacy (conversations already donated are deleted). They are used only to train Calogera, nothing else: the most common personal data is removed first and a person reviews them. People who use Calogera without an account donate nothing.
  • Licence: Calogera models are distributed under the Apache 2.0 licence (apache.org/licenses/LICENSE-2.0). Calogera-0.0 is at huggingface.co/Arrisx/Calogera-0.0-GGUF, with the NOTICE file of attributions. Calogera-1.0 uses the Arrisx/Calogera-1.0-lora adapter on top of Qwen3.5-27B: the Apache 2.0 licence text and Qwen’s attributions apply to it too.
  • Open-source components: Calogera Vista uses Florence-2-base (Microsoft, MIT licence) and Tesseract (Apache 2.0 licence); Calogera Voce uses Whisper (OpenAI, MIT licence); both in the ONNX version by onnx-community, with Transformers.js (Hugging Face, Apache 2.0).
  • “Qwen” is a trademark of Alibaba Cloud. Calogera is an independent project, not affiliated with or endorsed by Alibaba Cloud, Microsoft, OpenAI, Cohere Labs, OpenAssistant, LAION, Databricks or the Wikimedia Foundation.
  • Calogera was not built from scratch: each version starts from an open model (Qwen3.5) and was trained by Trynakria.