912c0f527e
Crawled https://disobey.dev/ and recreated all 10 blog posts with: - Preserved URLs using slug parameter in front matter - Original publication dates maintained - Proper Hugo front matter with tags/categories - Markdown formatting Posts migrated (chronological): - Github no more? (2025-08-17) - A true believer (2025-08-14) - how to post videos everyone can watch (2025-05-22) - wordpress docker setup (2025-05-05) - I built timeconvert.org (2025-01-31) - cryptoanarchy.info (2024-11-29) - quickly accessing llama3.2 from a terminal (2024-10-18) - The Untold Story of the Greatest Crypto Project Ever (2024-10-03) - the directory guy (2024-09-29) - AI blogs and resource (2024-09-26)
1.4 KiB
title, date, slug, tags, draft
| title | date | slug | tags | draft | ||
|---|---|---|---|---|---|---|
| quickly accessing llama3.2 from a terminal (or any other model) | 2024-10-18 | accessing-llm-through-cli |
|
false |
The author describes their workflow for integrating large language models into terminal-based work. Rather than switching to a browser or relying on a Telegram bot, they set up a more seamless solution using a remote GPU instance.
Technical Setup
The solution involves three key components:
- Remote Infrastructure: A cloud GPU instance running Ollama, accessible via SSH port forwarding (port 11434)
- Web Interface Option: Open WebUI for browser-based access
- CLI Tool: The
llmpackage with Ollama plugin support
Installation steps are straightforward:
- Install the main CLI tool
- Add Ollama plugin support
- Set Llama3.2 as the default model
Usage Examples
Simple Query: The tool handles basic prompts directly from the command line with natural conversational responses.
System Prompts: Users can pipe code or text through the tool with custom instructions, such as requesting code explanations.
Chat Mode: Interactive sessions allow extended conversations, though the author finds this less practical than web UI alternatives.
Limitations
The author notes the model includes safety guidelines that prevent certain types of content, and suggests exploring uncensored alternatives for less restricted interactions.