Software Requirements Classifier (LLM-powered via Ollama)

This is a command-line tool to analyze and classify software requirements using a local LLM served via Ollama. It supports both structured files and free-text input, and classifies requirements into categories like:

• Functional
• Non-Functional
• Performance
• Design
• Maintenance
• Ambiguous, Incomplete, etc.

Ideal for embedded systems, product specs, and agile teams that want AI-powered requirements triage 🔍⚡

🚀 Features

• 📂 Input via .txt, .pdf, or raw string
• 📤 Output to:
  • Console
  • PDF
  • JSON
• 🧠 Powered by local LLMs via ollama
• 💡 Built-in support for models like codellama:instruct
• 🐳 Docker-ready: prebuilt image with model loaded
• 🧪 Extensible and testable Python code

📦 Local Setup (Python CLI)

For developers who want to run the CLI directly

1. Clone the repo

git clone https://github.com/yourusername/software-requirements-classifier.git
cd software-requirements-classifier

2. Set up Python environment

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
pip install .

3. Run the CLI

reqclass --input "The system must respond in under 2 seconds." --console

reqclass --input path/to/reqs.txt --output report.pdf --output-json results.json

🐳 Docker Setup (LLM Backend)

This project uses a custom ollama Docker image with the LLM model preloaded.

1. Build Docker Image (If you want to customize)

docker compose build

Note: This step is optional. Please make sure you change gajesh to your Docker Hub username in the docker-compose.yml file if you want to push the image.

Or pull from Docker Hub (if you did not build the image locally):

docker pull gajesh/ollama-requirements-classifier

2. Run LLM server

docker compose up -d

This exposes the LLM at http://localhost:11434

📘 Usage Examples

From file:

reqclass --input tests/test_0_thermostat.txt --output report.pdf --output-json results.json

From raw text:

reqclass --input "The thermostat should log humidity every 10 minutes." --console

⚠️ Limitations

• 🧠 This tool relies on open-source LLMs like codellama:instruct
• LLM output may sometimes be:
  • Vague
  • Over-generalized
  • Sensitive to phrasing
• Strict formatting is enforced, but not 100% guaranteed (due to model limitations)
• Does not currently support batch processing of multiple documents

👤 Author

Made with ❤️ by Gajesh Bhat

🪪 License

MIT License — free to use, fork, and build on.