# Jacar

> Notas de producción sobre tecnología, DevOps e IA desde Madrid.

Technical blog by Javier Cañete, CEO of Jacar Systems. Covers DevOps, applied AI, Kubernetes, observability, cybersecurity, and product engineering. Content is published in Spanish and English.

## Key pages

- [Sobre — Javier Cañete (ES)](https://jacar.es/sobre/): About the site and its author, in Spanish.
- [About — Javier Cañete (EN)](https://jacar.es/en/about/): About the site and its author, in English.

## Recent posts

- [How to install OpenClaw with Docker and a local model](https://jacar.es/en/how-to-install-openclaw-with-docker/): How to install OpenClaw with Docker Compose, connect it to a local model with llama-server or Ollama, and harden the gateway: ports, token and skills.
- [Cómo instalar OpenClaw con Docker y un modelo local](https://jacar.es/como-instalar-openclaw-con-docker/): Guía para instalar OpenClaw con Docker Compose, conectarlo a un modelo local con llama-server u Ollama y bastionar el gateway: puertos, token y skills.
- [How to import a Hugging Face model into Ollama 0.34](https://jacar.es/en/import-hugging-face-model-into-ollama/): How to import a Hugging Face model into Ollama 0.34.1: convert safetensors to GGUF with llama.cpp, quantize to Q4_K_M and build it from a Modelfile.
- [Cómo importar un modelo de Hugging Face en Ollama 0.34](https://jacar.es/importar-modelo-hugging-face-en-ollama/): Cómo importar un modelo de Hugging Face en Ollama 0.34.1: convierte safetensors a GGUF con llama.cpp, cuantiza a Q4_K_M y créalo con un Modelfile.
- [How to upgrade Jellyfin 12 with Docker and keep a way back](https://jacar.es/en/upgrade-jellyfin-12-with-docker/): How to upgrade Jellyfin 12 in Docker from 10.11: /config backup, plugins, a 3.7 s migration, the required scan and a tested rollback.
- [Cómo actualizar Jellyfin 12 con Docker y poder volver atrás](https://jacar.es/actualizar-jellyfin-12-con-docker/): Cómo actualizar Jellyfin 12 en Docker desde la 10.11: copia de /config, complementos, migración de 3,7 s, escaneo obligatorio y vuelta atrás probada.
- [How to migrate from MinIO to Garage with Docker](https://jacar.es/en/migrate-minio-to-garage-with-docker/): How to migrate from MinIO to Garage v2.4.1 with Docker and rclone: 9,996 objects copied in 42 s, verification, S3 region, restic and what Garage lacks.
- [Cómo migrar de MinIO a Garage con Docker](https://jacar.es/migrar-minio-a-garage-con-docker/): Cómo migrar de MinIO a Garage v2.4.1 con Docker y rclone: 9996 objetos copiados en 42 s, verificación, región S3, restic y lo que Garage no soporta.
- [How to install llama.cpp on Linux, macOS and Docker](https://jacar.es/en/how-to-install-llama-cpp/): How to install llama.cpp 0.4.1 on Linux, macOS or Docker: the llama.app installer, Homebrew, the server image and CMake, with a GGUF served as an OpenAI API.
- [Cómo instalar llama.cpp en Linux, macOS y Docker](https://jacar.es/como-instalar-llama-cpp/): Cómo instalar llama.cpp 0.4.1 en Linux, macOS o Docker: instalador de llama.app, Homebrew, imagen server y CMake, con un GGUF servido como API OpenAI.
- [How to install Forgejo with Docker, Caddy and Forgejo Runner](https://jacar.es/en/how-to-install-forgejo-with-docker/): How to install Forgejo with Docker behind Caddy, register Forgejo Runner 13 and stay on the patched 16.0.4 release. Tested on arm64 with measured RAM.
- [Cómo instalar Forgejo con Docker, Caddy y Forgejo Runner](https://jacar.es/como-instalar-forgejo-con-docker/): Cómo instalar Forgejo con Docker detrás de Caddy, registrar Forgejo Runner 13 y quedarte en la versión parcheada 16.0.4. Probado en arm64 con RAM medida.
- [What the TrueNAS Proxmox plugin does and when it pays off](https://jacar.es/en/truenas-proxmox-plugin/): What the TrueNAS Proxmox plugin does: zvols over iSCSI or NVMe/TCP via the API, versus ZFS over iSCSI, NFS and Ceph, installation and limits.
- [Qué hace el plugin de TrueNAS para Proxmox y cuándo compensa](https://jacar.es/plugin-truenas-proxmox/): Qué hace el plugin de TrueNAS para Proxmox VE: zvols por iSCSI o NVMe/TCP vía API, frente a ZFS over iSCSI, NFS y Ceph, instalación y límites.
- [What Colibri is and how it runs giant MoE models by streaming experts from disk](https://jacar.es/en/colibri-giant-moe-models-from-disk/): What Colibri is, the C engine for Colibri MoE inference that streams experts from SSD: how it works, the speed to expect and what we measured on ARM64.
- [Qué es Colibri y cómo ejecuta modelos MoE gigantes leyendo expertos del disco](https://jacar.es/colibri-modelos-moe-gigantes-desde-disco/): Qué es Colibri, el motor en C que ejecuta modelos MoE gigantes leyendo expertos del SSD: cómo funciona, qué velocidad da y qué medimos en ARM64.
- [Engram: persistent memory for coding agents](https://jacar.es/en/engram-persistent-memory-coding-agents/): Engram brings persistent memory for coding agents with SQLite and MCP: what it does, its real limits, and when it beats mem0, Letta or Zep.
- [Engram: memoria persistente para agentes de código](https://jacar.es/engram-memoria-persistente-agentes-codigo/): Engram aporta memoria persistente para agentes de código con SQLite y MCP: qué hace, sus límites y cuándo compensa frente a mem0, Letta o Zep.
- [Activepieces and Windmill against n8n: which self-hosted automation to choose](https://jacar.es/en/activepieces-windmill-vs-n8n-comparison/): n8n alternatives compared by reading the licences: Activepieces ships MIT, Windmill mixes AGPLv3 with a proprietary build, and n8n forbids reselling it.
- [Activepieces y Windmill frente a n8n: qué automatización self-hosted elegir](https://jacar.es/activepieces-windmill-vs-n8n/): Alternativas a n8n comparadas leyendo las licencias: Activepieces es MIT, Windmill mezcla AGPLv3 y binario propietario, y n8n prohíbe revenderlo.