+==================================================================================+
| /\ ------- /\ |
| | | |
| | - - | |
| | | |
| | --- | |
| ------------- |
| | Horrible | |
| ------------- |
| MINI CODING AGENT |
+----------------------------------------------------------------------------------+
| |
| WORKSPACE C:\Users\Horrible\Code\AI-RESEARCH\horribleagent |
| MODEL qwen3.5:4b BRANCH main |
| APPROVAL ask SESSION 20260404-102010-f04466 |
| |
+==================================================================================+
HorribleAgent is a mini coding agent designed to assist developers in writing code efficiently. It leverages the power of large language models to provide code suggestions, automate repetitive tasks, and enhance the overall coding experience. Whether you're a beginner or an experienced developer, HorribleAgent can help you write better code faster.
- Code Suggestions: Get real-time code suggestions based on your current context.
- Automated Tasks: Automate repetitive coding tasks to save time.
- Multi-language Support: Supports multiple programming languages for a versatile coding experience.
- Local First: Runs entirely on your machine using Ollama.
- Python 3.12+
- Ollama: Download and install from ollama.com.
- Models: Pull a compatible model (e.g.,
ollama pull qwen3.5:4b).
To install HorribleAgent, follow these steps:
- Clone the repository:
git clone https://github.com/constcorrectness/horribleagent.git
cd horribleagent- (Optional) Create a virtual environment:
python -m venv .venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate- Install the required dependencies:
pip install -e .Alternatively, if you use uv:
uv syncStart the agent by running:
python main.pyOr with uv:
uv run main.pyYou can also provide a prompt directly from the command line:
python main.py "Refactor the current directory into a clean structure"HorribleAgent supports several command-line arguments:
| Argument | Default | Description |
|---|---|---|
--model |
qwen3.5:4b |
Ollama model name. |
--host |
http://127.0.0.1:11434 |
Ollama server URL. |
--approval |
ask |
Approval policy for risky tools (ask, auto, never). |
--max-steps |
6 |
Maximum tool/model iterations per request. |
--max-new-tokens |
512 |
Maximum model output tokens per step. |
--temperature |
0.2 |
Sampling temperature. |
While in the interactive shell, you can use the following commands:
/help: Show the help message./memory: Show the agent's distilled working memory./session: Show the path to the current saved session file./reset: Clear the current session history and memory./exit: Exit the agent.
This project is licensed under the GNU General Public License v3.0 (GPLv3). See the LICENSE file for details.