← ALL_LOGS

A Resource-Conscious Blueprint for Ubuntu (8GB RAM Rig)

DATE: 07/18/2026// 5 MIN READ// SYSTEMS_DOMAIN: HOW TO

Architecture Overview & Division of Labor

To prevent memory bottlenecks, we divide tasks between two core applications running within a unified interface:

  1. IDE (VSCodium): A telemetry-free, lightweight fork of VS Code. It serves as your visual workspace.
  2. AI Chat & Inline Assistant (Continue.dev): A sidebar extension used for codebase explanations, error debugging, and inline formatting edits.
  3. Agentic Layer (Aider CLI): A terminal-based pair programmer executing from within the VSCodium embedded panel. It drives multi-file structural edits and manages automated safe Git rollbacks.
  4. Cloud Brain (Google AI Studio - Gemini API): Handles all processing off-device, consuming 0MB of local RAM for LLM execution with a generous free tier of up to 1,500 requests per day.
+-------------------------------------------------------------+
|                     VSCODIUM IDE (Local)                    |
|                                                             |
|  +--------------------+  +-------------------------------+  |
|  |  Continue.dev      |  |  Source Code Workspace        |  |
|  |  Sidebar Panel     |  |  (Live File Updates)          |  |
|  |  [Explains / Ask]  |  |                               |  |
|  +--------------------+  +-------------------------------+  |
|                          |  Embedded Terminal Panel      |  |
|                          |  [Aider CLI Agent]            |  |
|                          +-------------------------------+  |
+-----------------------------------------+-------------------+
                                          |
                                 (Secure Cloud APIs)
                                          v
                          +-------------------------------+
                          |    Google AI Studio Cloud     |
                          |   [Gemini 2.5/3.1 Flash]       |
                          +-------------------------------+

Step 1: System Level Optimization (Ubuntu Only)

Ubuntu aggressively shifts inactive tasks into virtual swap files by default. For an 8GB RAM development workflow, reduce the operating system’s “swappiness” to prioritize physical memory execution.

  1. Open a standard terminal and check your default swappiness index: 
    cat /proc/sys/vm/swappiness
  2. Reduce the index to 15 to prevent background disk stutters: 
    sudo sysctl vm.swappiness=15
  3. To lock this configuration permanently across system reboots, append it to your system configurations: 
    sudo sh -c "echo 'vm.swappiness=15' >> /etc/sysctl.conf"

Step 2: Install VSCodium & The Continue Extension

  1. Install VSCodium cleanly via snaps to ensure automatic sandbox management: 
    sudo snap install codium --classic
  2. Launch VSCodium. Open the Extensions Marketplace panel (Ctrl + Shift + X).
  3. Search for Continue and click Install.
  4. Optional Optimization: Keep additional extensions disabled to maintain a footprint below 400MB RAM.

Step 3: Install Aider via Isolated Virtual Environments

To comply with Ubuntu’s PEP 668 policies and protect system packages, install Aider inside an isolated pipeline via pipx.

  1. Run the system dependency installations: 
    sudo apt update
sudo apt install pipx git htop -y
  2. Inject pipx application routes into your global system path configuration: 
    pipx ensurepath
  3. CRITICAL: Close your active terminal window completely and open a fresh terminal tab to initialize the new paths.
  4. Execute the isolated deployment of Aider: 
    pipx install aider-

Step 4: Provision Cloud Brain & Automate Global Configuration

1. Persistent API Key Injection

  1. Authenticate at Google AI Studio using a standard Google profile.
  2. Select Create API Key and clone the cryptographic string.
  3. Execute this command to inject it directly into your global Ubuntu profile: 
    echo 'export GEMINI_API_KEY="your-copied-api-key-here"' >> ~/.bashrc
  4. Refresh your environment states immediately: 
    source ~/.bashrc

2. Global Defaults Configuration

To avoid adding runtime flags manually during every session, define your preferences in a home-directory configuration file:

  1. Create a global Aider schema file: 
    nano ~/.aider.conf.yml
  2. Add this configuration to enforce the choice of your free cloud model: 
    model: gemini/gemini-2.5-flash
  3. Save and close the editor (Ctrl + O, then Enter, then Ctrl + X).

Step 5: Day-to-Day Development Workflow

Now that configuration is finalized, using your AI environment across any folder requires minimal overhead:

1. Workspace Activation

  1. Launch VSCodium and open any directory workspace (File > Open Folder).
  2. Drop down the integrated terminal block using Ctrl + ~.
  3. Rule: Aider depends on explicit version history tracking. If this is a new folder, instantiate a repository baseline once: 
    git init

2. Booting up the Agent

Type a single command to spin up your agent:

aider

Aider automatically ingests your global configs, accesses your system key, initializes communication with Gemini Cloud, and stands by.

3. Execution Commands for Quick Reference

Use these native shortcut protocols inside the terminal to direct your agentic pairs:

  • Target Specific Context Files:

    /add src/components/Auth.tsx

  • Prompting Code Modification Tasks:

    Refactor user session storage to preserve tokens securely across browser reloads.

  • Drop a File Out of Context Memory:

    /drop src/components/OldComponent.tsx

  • Safely Undo Mistaken AI Refactors: If the agent generates code that degrades execution parameters or breaks your layout, use the roll-back mechanism to instantly undo the changes:

    /undo

System Health Diagnostics Checklist

  • Checking RAM Overhead: Run htop in an external shell terminal. Monitor that system load stays low.
  • Watching Rate Quotas: Track daily usage metrics on your AI Studio Dashboard. Your free limit provides 1,500 operations every 24 hours.
  • Avoid Local Models: Never use local Ollama or embedding models on your local system, as this will consume your 8GB RAM instantly. Keep models remote with open(“ai_coding_setup_guide.md”, “w”) as f: 
    f.write(markdown_content.strip())

print("File written successfully.")

Quick Overview of what’s documented inside the file:

  1. Ubuntu Tuning: The system-level tweak to adjust vm.swappiness=15 so that your physical 8GB RAM takes full precedence over heavy disk-swapping when compilation or multiple workspaces run simultaneously.
  2. Pipx Isolation Setup: Clean, PEP 668 compliant global installations for Aider.
  3. Automated Configuration Configuration: Bypassing tedious repetitive flags by dropping a permanent .aider.conf.yml definition straight into your Linux home (~) directory.
  4. Division of Labor Workflow: A breakdown of when to look at the left sidebar panel (Continue.dev) for questions versus using the bottom panel (Ctrl + ~) to let Aider run multi-file agentic iterations on your workspace code using your daily 1,500 free tokens.

You can save this file right into your main learning directory or notes base for easy reference whenever you initialize a brand new folder. Let me know if you want to customize any part of this layout as you begin coding!