📃 Log

This article connects from Raspberry Pi to Ollama running on the PochomLab machine inside the LAN, and confirms that Python can send requests to the local LLM.

This article sets up a Python virtual environment on Raspberry Pi, installs requests, checks basic operation, and creates a requirements.txt file.

A record of connecting to a Raspberry Pi from VS Code on a work PC using Remote-SSH, then editing files and running Python on the Raspberry Pi.

A record of the initial setup for Raspberry Pi 4, including SSH connection, package updates, remote desktop setup with xrdp, and HDMI resolution and audio adjustments.

A record of creating an installation USB drive with Raspberry Pi Imager, booting from it, and writing Raspberry Pi OS to the SSD.

A record of preparing a Raspberry Pi 4 and its peripherals, assembling it into an Argon ONE M.2 case, and checking the monitor setup.

I installed Open WebUI, connected it to Ollama, and confirmed that I could use my local LLM from another PC on the same LAN.

I installed Ollama on Windows and confirmed that a local LLM could actually run by testing a lightweight model.

For my local LLM setup, I asked several generative AIs to suggest model candidates for two purposes: Raspberry Pi connection experiments and ZINE writing support. Then I narrowed down which models to try first.

As a first step toward adopting local LLMs, I organized basic terms such as Ollama, LM Studio, model naming, quantization, and RAG.

We began recomposing the original Japanese manuscript and integrating English text for the English edition of Pochomkin Egaki-sho. By switching to a layout where text is placed over completed artwork, we are building a framework that will make future multilingual editions easier to create.

The KDP paperback proof copy of the English edition of 'Egaki-sho: Let’s Create! ZINE — The Creation Process' has arrived. This log records the revised cover results and the layout impressions that became clear in print.

A sample that organizes the specifications and writing rules for Log post front matter.

The first KDP paperback proof copy of 'Egaki-sho: Let’s Create! ZINE — The Creation Process' has arrived. This log records what became clear after seeing the physical copy, especially the inner margin and cover layout.

A build log of the local generative AI workstation used in PochomLab. This setup is centered around an RTX 4070 SUPER and 64GB of memory, designed for practical long-term operation.

Implemented Google Analytics 4 with Consent Mode. Analytics events are sent only after user consent.

The “Great Encyclopedia of Pochomkin” generated by Google’s AI search. A record of the moment when the world connected after 26 years.

Separated the AstroWind site into a multilingual structure using /ja and /en. Updated navigation, layout, canonical, and hreflang settings.

Modified the AstroWind header navigation to use anchor scrolling. Disabled ClientRouter and implemented in-page navigation.

Organized the frontmatter specification and writing rules for event posts. Clarified type classifications and display behavior, and prepared navigation to the Events page.

Set up the initial PochomLab site environment based on AstroWind. Organized the base configuration and development setup.