Una GPU VPS usa e getta per i tuoi LLM: LiteLLM, Ollama, vLLM e storage cifrato
Noleggia una GPU VPS, trasformala in una API LLM compatibile OpenAI (LiteLLM, Ollama, vLLM), poi distruggi la VM: modelli e dati vivono su un volume LUKS cifrato.
Noleggia una GPU VPS, trasformala in una API LLM compatibile OpenAI (LiteLLM, Ollama, vLLM), poi distruggi la VM: modelli e dati vivono su un volume LUKS cifrato.
Rent a GPU VPS, turn it into an OpenAI-compatible LLM API (LiteLLM, Ollama, vLLM), then destroy the VM — your models and data survive on an encrypted LUKS volume.
How I came to understand language models a little better by trying to squeeze one into a few megabytes.
Come ho capito un po' meglio i modelli linguistici provando a farne stare uno in pochi megabyte.
A browser-first workbench for Claude Code, Codex and friends: one Docker Compose stack, SSO+2FA on every door, persistent terminal sessions you can rejoin from anywhere — including the queue at the post office.
A plain-language LLM glossary — but not alphabetical: in dependency order, so by the time you reach Mixture of Experts you already have all the pieces. With analogies for sysadmins.
What an MCP server actually is (with a detour through Roman roads and railway gauges), the Mataken use case, and an honest, reservations-included argument for making your content legible to agents — not just to browsers.
Build a complete self-hosted LLM environment: GPU VM → Ollama → LiteLLM proxy → Nginx → Continue.dev in VS Code. Includes all the integration steps and config details to get everything working seamlessly.
Can AI recognize faces, detect if someone’s real, and build profiles — all in real time? In this proof-of-concept, we combine YOLO, InsightFace, and blink detection to create an intelligent system that watches streams so humans don’t have to. Built with Python, open-source tools, and curiosity.
A fun and practical project that uses Python and AI to classify SMS replies like “Yes”, “Maybe later”, or “Stop”. From keywords to embeddings to GPT, follow the journey of building a multilingual intent classifier that goes way beyond Excel hacks.
In this post, we explore how to build a smart travel assistant using the OpenAI Agents SDK. With just a few tools and guardrails, we created an agent that fetches weather, finds places to visit, and gives real-time travel tips — all in one Markdown-friendly reply.