Argus is a local-first observability tool for LLM calls. It logs every prompt you send to a local Ollama model (or OpenAI) — latency, token counts, estimated cost, success or failure, and the machine's CPU and memory during the call — into a local SQLite database, then visualizes it all in a dashboard. No cloud, no accounts, no cost.
Highlights
-
Automatic logging of every call through a thin client wrapper.
-
Latency, token, and cost metrics with system resource snapshots.
-
FastAPI backend over SQLite, with a Streamlit dashboard.
-
Runs entirely on your own machine.
Stack
Python, FastAPI, Streamlit, SQLite, and Ollama or OpenAI.
Screenshots
