This thesis presents a chat-based agent that enables natural language interaction with the Linux command line. The system uses a lightweight local language model to translate user queries into Bash commands. Generated commands are executed in a sandboxed environment, with an iterative feedback loop to improve accuracy. The approach is evaluated using execution-based and usability metrics. Results show that local AI agents can effectively support command-line interaction while maintaining safety and efficiency.
This thesis presents a chat-based agent that enables natural language interaction with the Linux command line. The system uses a lightweight local language model to translate user queries into Bash commands. Generated commands are executed in a sandboxed environment, with an iterative feedback loop to improve accuracy. The approach is evaluated using execution-based and usability metrics. Results show that local AI agents can effectively support command-line interaction while maintaining safety and efficiency.
Linux Command Line Agent: Chat-Based Interaction for Terminal Commands.
NAYYERI, SEYYEDEH REZVAN
2025/2026
Abstract
This thesis presents a chat-based agent that enables natural language interaction with the Linux command line. The system uses a lightweight local language model to translate user queries into Bash commands. Generated commands are executed in a sandboxed environment, with an iterative feedback loop to improve accuracy. The approach is evaluated using execution-based and usability metrics. Results show that local AI agents can effectively support command-line interaction while maintaining safety and efficiency.| File | Dimensione | Formato | |
|---|---|---|---|
|
Nayyeri_SeyyedehRezvan.pdf
accesso aperto
Dimensione
813.48 kB
Formato
Adobe PDF
|
813.48 kB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/116605