This week was very hot because Google published Gemma, a smal-size LLM, and also Nvidia has a rush on stock market, with a evaluation around 2000 BILLIONS dollars. Last but not least, Mistral.ai published a service for its more advanced model, which can also run on-prem.

Read More

Sometimes you need to fix a production defect. To do it, you need to mirror a specific set of data, and do a test to proof the bug and then fix it.

Read More

We already discussed about Queue management solutions in the past, and I am always happy to write about it. Queue managers are not easy to implement, and there is a reson if IBM MQ Series is still a successful product. Some month ago, a big bank customer asked me to provide a small queue implementation to increase asynchronous internal processing of our payment solution.

The project had very strong contraints: I could not use existing queue system because they were not yet available, and I need to be able to provide microservice parallelism in a cloud-environent.

I have very little time to provide a solid solution, and re-inventing the wheel was not an option. Performance was important, but we plan to have a managable numbers of transactions per seconds, far behind modern cloud database capacity.

Challenge accepted.

Read More

I want to share with you two tools I use a lot on some situation, where I need to write short manuals for co-worker or end-customers.

Read More

Good Karma Kit

The Good Karma Kit is “a Docker Compose project to run on servers with spare CPU, disk, and bandwidth.” I like the idea in principle, but it is always a complex thing to do, because if you host unknown content, you can get in trouble easily (like pirated content or worst…)

Read More

Anubis

Weigh the soul of incoming HTTP requests using proof-of-work to stop AI crawler

LLM web spiders are creating troubles to tiny website, because they seems to ignore robots.txt directive and scrape all the website at the bild search of information to ingest.

Read More