https://github.com/hweichelt/explaidllm
CLI tool for debugging clingo using a combination of MUS and LLMs
https://github.com/hweichelt/explaidllm
answer-set-programming debugging llm mus
Last synced: 24 days ago
JSON representation
CLI tool for debugging clingo using a combination of MUS and LLMs
- Host: GitHub
- URL: https://github.com/hweichelt/explaidllm
- Owner: hweichelt
- License: mit
- Created: 2025-07-01T11:57:29.000Z (10 months ago)
- Default Branch: main
- Last Pushed: 2026-01-23T18:04:22.000Z (3 months ago)
- Last Synced: 2026-01-24T08:23:39.593Z (3 months ago)
- Topics: answer-set-programming, debugging, llm, mus
- Language: Python
- Homepage:
- Size: 177 KB
- Stars: 0
- Watchers: 0
- Forks: 1
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# ExplaidLLM
ExplaidLLM is a command line tool aimed for debugging clingo programs. This is achieved by combining the Minimal
Unsatisfiable Subset (MUS) functionalities of `clingexplaid` together with the natural language capabilities of an LLM.
## Installation
### From Source
```bash
pip install .
```
### From PyPi
```bash
pip install explaidllm
```
## Configuration
> [!NOTE]
> Currently ExplaidLLM only supports the OpenAI LLM API. The implementation of other API's can be requested through Issues.
### LLM API
Before using ExplaidLLM you need to store your API Key to prompt your LLM of choice.
You can do this in two different ways:
#### Using a `.env` file
+ Recommended for installation from source
```.dotenv
OPENAI_API_KEY=
```
#### Using the `-k` API-KEY option
+ Recommended for installation from PYPI
```bash
explaidllm examples/test.lp -k=$YOUR_OPENAI_KEY_ENV_VAR
```
## Usage
ExplaidLLM allows you to quickly debug your unsatisfiable ASP Programs. For the default behaviour just use this command:
```bash
explaidllm example/test.lp
```
The default behaviour converts all you program atoms to assumptions, computes an MUS (Minimal Unsatisfiable Subset) from
them, finds a matching unsatisfiable constraint and then prompts the configured LLM for an explanation.
### Assumption Signatures
If you want to filter only certain assumption signatures use the option `-a` to specify the signatures to be converted.
```bash
explaidllm example/test.lp -a x/1
```
### LLM Models
You can also specify which LLM Model should be used for the explanation using the `-m` model option.
A list of available models can be found on the `--help` page.
```bash
explaidllm example/test.lp -m='gpt-4o'
```