Only human made additions are the images.
To get started with this project, follow these steps:
git clone g-hano/ReadmeCreator
pip install -r requirements.txtThis project uses a combination of natural language processing (NLP) and graph-based workflows to generate a README.md file based on code summaries. The system prompts users to input a directory name, list of files, and then creates summaries for each file using an LLM model.
To use this project, simply follow the prompts in the main.py script:
- Enter a directory name.
- The system will generate summaries for each file based on the LLM model.
- The system will decide which summaries to re-write
- You can use README.md for your projects.
- Project Lead: M.Cihan Yalçın
This project is licensed under the MIT License.
The graph-based workflow is defined in the graph.py script. It uses a StateGraph to represent the workflow, which consists of nodes and edges. The nodes represent different steps in the process, while the edges represent the relationships between them.
get_directory: Asks the user for a directory name.get_all_files: Lists all files in the specified directory.get_summaries: Generates summaries for each file using the LLM model.judge: Evaluates the generated summaries and decides which ones to rewrite or include in the README.md file.rewrite: Rewrites selected summaries based on the LLM model.readme: Creates a README.md file based on the generated summaries.
The edges represent the relationships between nodes. For example, the edge from get_directory to get_all_files means that after getting the directory name, the system will list all files in that directory.
The LLM model used for generating summaries is defined in the agents.py script. It uses an Ollama instance with a specific model and prompt.
The system prompts are defined in the configs.py script. They include the LLM model, summary folder, system prompt folder, and LLM prompt folder.
I hope this README.md file meets your requirements! Let me know if you need any further assistance.