SimiNotes - CLI Tool for Similar Note Retrieval

SimiNotes is a command-line interface (CLI) tool written in Python that enables users to discover similar notes within their notes collection. The tool utilizes sentence embeddings with sbert to compare a given query against a corpus of user notes.

Features

• Embedding: Utilizes sbert to generate vector embeddings for notes.
• Similarity Search: Finds notes similar to a given query based on embeddings.
• Configurable: Allows users to configure directories, file extensions, and exclusion criteria.

Installation

Before using this CLI tool, ensure that you have Python and pip installed. Additionally, install the PyTorch library by following the steps below:

1. Download and install Python and pip from the official site.

2. Install the PyTorch library separately based on your requirements. For the CPU version, use the following command:

   pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

Now, install the SimiNotes CLI tool:

pip install siminotes

Configuration

Before using the CLI, configure some essential values, such as the notes directory and exclusions. SimiNotes uses a configuration file (config.txt) to set preferences. Configure the tool by creating the file in the appropriate configuration directory.

Configuration Directory

• Linux:

~/.config/siminotesconfig/

• macOS:

~/.siminotesconfig/

• Windows:

For Windows, use AppData\Roaming for per-user configuration:

AppData\Roaming\Siminotes\

Alternatively, place config.txt in the home directory:

~/.siminotesconfig/

Configuration File (config.txt)

Create a config.txt file in the configuration directory. Below is an example configuration:

notes_dir = /path/to/your/notes
exclude_dir = directory1,directory2
exclude_file = file1,file2
note_extension = .md

Configuration Parameters:

• notes_dir: Path to the directory containing your notes.

• exclude_dir: Comma-separated directories to exclude from the search. Paths should be relative to notes_dir.

• exclude_file: Comma-separated files to exclude from the search. Paths should be relative to notes_dir.

• note_extension: The extension of your note files (e.g., .md).

Usage

Now let's use our cli,

Command-Line Arguments

• Query via Text:

siminotes text "Your Query Text"

• Query via File:

siminotes file filename

Both will result in,

Top files which are similar to given query:
Value range from -1 to 1, where going toward 1 means note is close to query

/... with score 0.43386968970298767

/... with score 0.42138463258743286

...

Troubleshooting

If you encounter any errors or problems with this tool, please open an issue in the repository.

License

This project is licensed under the MIT License.

Contributing

Feel free to contribute to SimiNotes by creating issues or submitting pull requests.

Acknowledgments

Sentence-BERT for sentence embeddings.

Future:

• I feel like simple dot product hits are enough to find similar notes but in future if there is need to improve results then consider this roop Retrieve and Rerank Vid Tut

• If it is taking more memory, then we can quantise the vectors into int8 Quantisation Guide Github Repo to check