https://github.com/mitdbg/logos
Human-in-the-Loop Causal Analysis of Log Files
https://github.com/mitdbg/logos
Last synced: 4 months ago
JSON representation
Human-in-the-Loop Causal Analysis of Log Files
- Host: GitHub
- URL: https://github.com/mitdbg/logos
- Owner: mitdbg
- License: mit
- Created: 2024-04-01T20:48:25.000Z (about 2 years ago)
- Default Branch: main
- Last Pushed: 2024-11-29T00:48:20.000Z (over 1 year ago)
- Last Synced: 2024-11-29T01:34:47.835Z (over 1 year ago)
- Language: Python
- Homepage: https://mitdbg.github.io/logos/
- Size: 6.17 MB
- Stars: 3
- Watchers: 9
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# LOGos
Utilizing system logs to perform causal analysis. You can access the documentation [here](https://mitdbg.github.io/logos).
Please begin by installing the Python packages required for this project by running `pip install -r requirements.txt`.
### OpenAI integration
In order to use the LLM-powered capabilites of LOGos, please add a `.env` file to the root of this repo and define `OPENAI_API_KEY` appropriately.
### Trying out LOGos
For an introduction to our Python-based interface, you can turn to our demo notebook at [`demo/demo.ipynb`](demo/demo.ipynb).
We also offer a simple UI built using [Streamlit](https://docs.streamlit.io/). You can launch it by running [`demo/run_ui_demo.sh`](demo/run_ui_demo.sh) and following the resulting URL.
### Reproducing our evaluation
To reproduce the evaluation from our VLDB paper, please follow the following steps:
1. Follow the instructions in `dataset_files/README.md` to gain access to our datasets.
2. Within `evaluation/`, you will find directories based on each experiment presented in our paper. Based on the experiment you would like to reproduce, switch into the appropriate directory and run the `reproduce.sh` script (you may need to edit file permissions to make it executable). This will run the experiment and plot the results.
3. Find the resulting plots in `evaluation/repro_plots/`. The raw data for each plot will be saved in `evaluation/repro_plots_data/`.