All projects
AI assistantin progress

asistenIT

An AI assistant for everyday IT tasks: answering questions, drafting scripts, and automating small chores from one chat interface.

PythonHugging FaceLLM

the problem

Switching between five different single-purpose CLI tools for everyday IT tasks -- quick scripts, explanations, lookups -- breaks focus. A single chat interface that can actually help across all of them is faster.

approach

A lightweight Python chat client against Hugging Face's hosted Inference API (Zephyr-7B-beta), with a system prompt explicitly tuned to treat follow-up questions as continuations of the same conversation instead of resetting context on every turn, and to stay on-topic rather than drifting.

architecture

Hugging Face Inference Client

Calls a hosted Zephyr-7B endpoint rather than self-hosting a model, keeping the client itself thin.

Context-aware system prompt

Explicitly instructed to treat a follow-up as continuing the prior answer's context, not a fresh, unrelated question.