RAG Notes

A solution to organize notes using Vector Databases offline.

You'll get downloaded from huggingface the BGEM3FlagModel, but you may choose between the ones supported by FlagEmbedding to embed the files into the database

Project Structure

The RAG Notes App is organized into the following directory structure:

RAG Notes
├── .git
├── .gitignore
├── LICENSE
├── __pycache__
├── app.py
├── document.py
├── embedding.py
├── requirements.txt
├── styles.py
└── vect_db.py

Important Files

• app.py: The main application file that initializes the GUI and manages user interaction.
• document.py: Contains the Document class for note representation and methods for file handling.
• embedding.py: Handles the embedding of text using the BGEM3FlagModel for document similarity.
• vect_db.py: Manages the vector database for storing, retrieving, and manipulating documents.
• styles.py: Defines visual styles and configurations for the application components.
• requirements.txt: Lists Python package dependencies required to run the application.

Installation

To install the required dependencies, create a virtual environment and run:

pip install -r requirements.txt

Usage

To run the application, execute:

python app.py

Features

• Note Management: Create, delete, and update notes.
• Search Functionality: Search for notes with auto-suggestions based on user input.
• Clipboard Monitoring: Automatically detect and save text copied to the clipboard as new notes.
• Local Document Embedding: Utilize the BGEM3FlagModel for semantic similarity between notes.
• Easy to implement a LLM: the vect_db.py allows CRUD operations quickly and stable

Classes and Functions

Document Class

class Document:
    def __init__(self, id: str, content: str):
        self.id = id
        self.content = content

    def to_file(self, file_path: str):
        ...

• Attributes:

  • id: Unique identifier for the document.
  • content: Text content of the document.

• Methods:

  • to_file(file_path): Saves the document content to a specified file.

Functions in vect_db.py

• add_doc(doc: Document) -> None: Adds a document to the vector database and updates the embeddings.
• delete_doc(doc_to_delete: Document) -> None: Deletes a specific document from the database.
• update_doc(doc: Document) -> None: Updates an existing document in the database.
• find_docs(query: str, k: int, min_similarity_threshold: float) -> List[Document]: Retrieves documents matching a query based on semantic similarity.

GUI Components

Main Application Class: NotesApp

class NotesApp(ctk.CTk):
    ...

• Initialization: Configures the main window, initializes GUI elements, and starts clipboard monitoring.

• Methods:

  • copy_note_content(): Copies the current note content to the clipboard.
  • load_notes(query=""): Loads notes based on a search query.
  • add_note(event=None): Handles note creation events.
  • search_notes_action(event=None): Executes search operations and displays the results.

Styles Configuration

The styles.py file manages the theme and user interface styling, including color palettes, fonts, and widget styles. Key styles include:

• COLOR_PALETTE: A dictionary that defines colors used throughout the application.
• APPEARANCE_MODE & DEFAULT_COLOR_THEME: Define the initial appearance settings.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

• CustomTkinter: A modern Tkinter wrapper used for better UI flexibility.
• BGEM3FlagModel: Used for obtaining document embeddings.