USB LLM Portable Vault v0.2.0

USB LLM Portable Vault v0.2.0

USB LLM Portable is an encrypted, no-install AI browser workspace that can
connect to online APIs or a local OpenAI-compatible LLM server.

New in this version

• Added IO-HO branding and logo
• Added “Developed by Antonio Gargiulo”
• Added links to io-ho.com and dragossido.com
• Added encrypted provider credential storage
• Added encrypted conversation storage on the USB
• Added temporary attachment processing
• Added separate launchers:
  • RUN_USB_LLM_ONLY.bat
  • RUN_USB_LLM_WITH_LOCAL_AI.bat
• Included the working local llamafile server package

How to use

Browser/API mode

Run:

RUN_USB_LLM_ONLY.bat

Browser plus local AI mode

Run:

RUN_USB_LLM_WITH_LOCAL_AI.bat

Security

Credentials and conversations are encrypted using the user's master password.

Original attached images, documents, audio and video files are not stored in
the portable workspace.

Requirements

• Windows
• A modern web browser
• Sufficient RAM for the selected local model
• Internet access only when using online providers

Developer

Developed by Antonio Gargiulo

• io-ho.com
• dragossido.com

USB LLM Portable Vault v0.1.0

USB LLM Portable v0.1.0

The first public version of USB LLM Portable introduces a lightweight,
no-install AI browser workspace designed to run directly from a USB drive.

Main features

• Portable browser-based AI interface
• No Node.js, npm, Rust or Visual Studio required
• Runs through a local PowerShell web server
• Connects to OpenAI-compatible APIs
• Supports local and online AI providers
• Provider settings for:
  • API base URL
  • Model name
  • API key
• Local and online privacy indicator
• Text conversation interface
• Image attachment support for compatible vision models
• Relative paths, allowing the folder to work from different USB drive letters
• No installation required on the host computer

How to use

1. Extract the package to a USB drive.

2. Run:

   START_USB_LLM.bat

3. The interface opens automatically in the default browser at:

   http://127.0.0.1:8090

4. Enter the details of an OpenAI-compatible provider.

Local provider example

Use these settings for a local compatible server:

• Base URL: http://127.0.0.1:8080/v1
• Model: local-model
• API key: leave empty if the local server does not require one

Compatible local servers may include:

• llama.cpp
• LM Studio
• LocalAI
• Ollama with an OpenAI-compatible endpoint
• Other compatible local inference servers

Online provider support

USB LLM Portable can connect to online services that provide an
OpenAI-compatible chat-completions API.

The application clearly displays whether the selected provider is:

• LOCAL
• ONLINE

This helps users understand when their messages may leave the computer.

Image support

Images can be attached to messages and sent using the common
OpenAI-compatible image_url format.

Image analysis requires a model and provider that support vision input.

Portability

The application uses relative paths, so the USB drive letter can change between
computers without requiring configuration changes.

The interface and server files remain on the USB drive.

Known limitations

• Conversation history is not yet stored permanently
• Provider credentials are not yet stored in an encrypted USB vault
• Full document, audio and video support is not yet implemented
• Local model files and local inference runtimes are not included
• Provider-specific integrations such as Gemini and Anthropic are not yet included
• Streaming responses are not yet supported

Security

The local web server listens only on:

127.0.0.1

This means the interface is intended to be accessible only from the same
computer.

Users should avoid storing API keys in plain-text files.

Platform

• Windows 10 or Windows 11
• PowerShell
• A modern web browser

Developer

Developed by Antonio Gargiulo

• io-ho.com
• dragossido.com