This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine
PHANTOM SHELL: A Local Gemma Investigator for the Ghost in Your Machine
What I Built
PHANTOM SHELL is a Windows-first diagnostic terminal for investigating mysterious system slowdowns. It gathers CPU, memory, disk, uptime, and bounded process information, then checks that evidence for common warning signs.
The “ghost” is the uncertainty around what a computer is doing. PHANTOM SHELL gives that uncertainty a traceable path: deterministic checks identify the observed signals, and a local Gemma model offers possible explanations grounded in collected evidence. Those explanations are hypotheses, not diagnoses. The fictional Story Mode is labeled as fiction.
Demo
phantom-shell-demo.mp4 - Google Drive
drive.google.com
The demo uses clearly labeled sample values; live process identifiers and machine-specific readings are not shown.
Code
How I Built It
PHANTOM SHELL is written in Python. It collects system telemetry, applies deterministic anomaly checks, and packages bounded evidence with IDs. The model runs through local Ollama using the already-installed gemma4:e4b model.
Python controls which read-only tools can run. Gemma can suggest an explanation, but it cannot choose arbitrary commands. The app checks that evidence references point to collected IDs and keeps model hypotheses separate from measured facts. Rich powers the terminal UI, and SQLite stores local history.
Why Does Open Innovation Matter?
Running Gemma locally means the diagnostic evidence can stay on the user’s machine instead of being sent to a hosted inference API. The model, prompts, evidence format, and validation path are available for people to inspect and adapt.
Open tooling also made it practical to use a model already available on the machine, without adding a new model download or relying on a per-request cloud service.
The Spirit of It
A process using unexpected resources can feel like something invisible is at work. PHANTOM SHELL makes that moment less mysterious: it shows what the machine measured, what Gemma suspects, and which evidence supports the suggestion. It leaves the final judgment with the person at the terminal.
Prize Categories
- Best Use of Gemma
Top comments (0)