HN Showcase
Back to browse
Open-Jet – self-hosted Agentic TUI for air-gapped Jetsons

Open-Jet – self-hosted Agentic TUI for air-gapped Jetsons

by lforster·Mar 2, 2026·2 points·2 comments
Visit ProjectView on HN

AI Analysis

●●●BangerWizardryBig BrainNiche Gem

Terminal agent for Jetson with memory-aware context windowing and TensorRT optimization.

Strengths
  • •Hardware-aware inference routing (llama.cpp, TensorRT-LLM) with automatic layer quantization detection avoids OOM on unified-memory Jetson boards
  • •Structured JSONL logging and approval gates for tool execution (shell, file ops) mean safe local automation without blind execution
  • •Tested across 6 Jetson SKUs (Nano through AGX Orin 64GB) with benchmarked performance (17 tok/s on Nano 8GB) proves real hardware optimization, not marketing
Weaknesses
  • •Audience is genuinely niche: requires Jetson hardware, Python 3.10+, and comfort with terminal interfaces; kv-cache tuning still incomplete
  • •Early alpha with UX gaps (no progress indicators, cache expiry not signaled, batch process toggle needs 2 clicks to restart)
Target Audience

ML engineers, edge device developers, air-gapped deployment teams

Post Description

Hi HN,

I am building a Terminal User Interface (like Claude Code) for self-hosted AI agents on Jetsons. Works in air-gapped environments. Unlike other solutions, this is optimised for unified memory machines, as to avoid OOM errors.

The agent can do stuff like edit, read, create files - manage and interpret data locally.

Currently, it gets ~17 tok/s on Jetson Orin Nano 8GB using Qwen3-4B-Instruct-4bit In the future, adding TensorRT .engine support which will boost inference further. I am trying to get the memory footprint down, so if anyone has knowledge on kv cache optimisation, that would be great.

I would love to get your feedback and people try running it on more capable devices and models - post your results here.

Run ``` pip install open-jet open-jet --setup ```

Webiste: https://www.openjet.dev/ Directly on Pypi: https://pypi.org/project/open-jet/ Repo: https://github.com/L-Forster/open-jet/

Similar Projects

AI/ML●●Solid

Quoracle: Self-replicating multi-LLM-consensus agents (Elixir)

Quoracle actually does something interesting: it queries a pool of models and only executes actions they agree on, while letting agents spawn children and persist full state to Postgres — all visible in a LiveView dashboard. The per-model conversation history, recursive hierarchy, and explicit consensus pipeline are clever touches; it’s clearly aimed at experimentation rather than drop-in production use (the README even flags security and deployment caveats).

WizardryNiche Gem
shelvick
213mo ago
AI/ML●●Solid

Routiium – self-hosted LLM gateway with a tool-result guard

Guards tool outputs against injection attacks, unlike LiteLLM or Helicone.

Solve My ProblemBig Brain
deadpixel
201mo ago