Tom

0Popularity

About Tom

Tom is a local AI assistant designed to work with files, ideas, and everyday tasks on a Windows computer. It runs on the user’s machine using Gemma 4 via llama.cpp and does not require a GPU. The application is distributed as a full offline installer or a small network installer that downloads components on first run. Tom displays visible work through ASCII states and a context ring, allowing users to inspect its progress and actions. It can read and create local files, summarize documents, transform tables, generate checklists, and inspect simple images. The assistant can also run programs and interact with supported browser pages with visible actions and review controls. Tom is positioned as a small, focused tool rather than a general-purpose AI, with limitations noted in browsing, large coding jobs, and long tasks. It is built on open-source components including Gemma 4, Node.js, SQLite, and Playwright, and is released under the Apache 2.0 license.

Key features

Local model inference using Gemma 4
File reading and creation
Visible ASCII states showing work progress
Context ring for usage tracking
Built-in help and review controls
Program execution with visible actions
Browser page interaction support
Apache 2.0 open-source license

Use cases

Summarizing documents or files
Transforming small tables or data
Generating checklists or simple image inspections

Pros

Runs locally on Windows without requiring a GPU
Open-source software with transparent components
Visible work and actions for user inspection
Supports file operations and focused tasks like summarization
No internet required for local model inference

Cons

Beta software without a publisher signature
Requires manual installation of Microsoft Visual C++ runtime
Limited to Windows 10/11 x64
No GPU acceleration supported

Frequently asked questions about Tom

What is Tom and what does it do?

Tom is a local AI assistant designed for Windows that works with files, ideas, and everyday tasks. It runs on the user's machine using Gemma 4 via llama.cpp without requiring a GPU, and displays its progress through visible ASCII states and a context ring.

Who is Tom suitable for?

Tom is suitable for users who need a small, focused AI assistant for local tasks such as summarizing documents, transforming tables, generating checklists, or inspecting simple images. It is not intended for large coding jobs, extensive browsing, or long tasks.

How do I install Tom?

Tom can be installed using either a full offline installer or a small network installer. Before installation, ensure Windows 10/11 x64 is running with a supported browser and Microsoft’s Visual C++ x64 runtime is installed. The full installer includes all components, while the network installer downloads them on first run.

Does Tom require an internet connection?

While Tom runs locally for most tasks, some features like browsing, search, downloads, and update checks require an internet connection. Local processing does not guarantee offline functionality for all features.

What are Tom’s limitations?

Tom is not designed for browsing, large coding jobs, or long tasks. Its reliability varies depending on the task, and users should verify important work as local processing does not ensure correctness. Some features, like webpage tasks, may fail.

How can I verify the integrity of the Tom installer?

The full installer is provided as an original EXE file with a SHA-256 checksum for verification. Users can download the checksums and compare them locally to ensure the installer’s integrity before running it.