https://github.com/nithin-sudarsan/vectorly
Train on example sentences, then query to find the most similar ones with similarity percentages.
https://github.com/nithin-sudarsan/vectorly
Last synced: about 2 months ago
JSON representation
Train on example sentences, then query to find the most similar ones with similarity percentages.
- Host: GitHub
- URL: https://github.com/nithin-sudarsan/vectorly
- Owner: nithin-sudarsan
- Created: 2025-07-24T10:31:48.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-07-26T10:10:00.000Z (2 months ago)
- Last Synced: 2025-07-26T16:32:21.371Z (2 months ago)
- Language: Python
- Size: 6.84 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Vectorly
A terminal-based interactive app to generate vector text-embeddings, that gets smarter as you use!
Train on example sentences, then query to find the most similar ones with similarity percentages.
---
## Features
- **Initialize a vector database** for storing phrase embeddings.
- **Feed new phrases** to the corpus and update the vector store.
- **Query phrases** to find the most similar ones in the corpus using cosine similarity.
- **View corpus, vocabulary, and similarity matrix**.
- **Clear or delete** the vector database.
---
## Installation
1. Clone the repository:
```sh
git clone https://github.com/nithin-sudarsan/vectorly.git
cd vectorly
```
2. Install dependencies:
```sh
pip install numpy pandas
```
---
## Usage
Run the app from the terminal:
```sh
python main.py
```
### Commands
- `init`
Initialize a new vector database.
- `feed`
Feed new phrases into the existing corpus interactively.
- `query`
Query a phrase to find the most similar ones from the corpus.
- `view [-c] [-v] [-m]`
View corpus (`-c`), vocabulary (`-v`), and similarity matrix (`-m`).
- `clear`
Clear all phrases from the existing corpus.
- `delete`
Delete the existing corpus.
- `--help`
Display help message.
---
## How It Works
- **Vector Representation:**
Each phrase is tokenized and represented as a vector based on word counts (bag-of-words model).
- **Cosine Similarity:**
Similarity between phrases is computed using cosine similarity.
- **Persistence:**
The corpus and vectors are stored in a binary file `.vectorstore` using Python's `pickle` module.
---
## File Structure
- `main.py`: CLI entry point and command parser.
- `DBOperations.py`: Handles database operations (init, feed, query, view, clear, delete).
- `VectorStore.py`: Implements the vector store, vectorization, similarity computation, and matrix viewing.
- `utils.py`: Utility functions (cosine similarity).
- `.vectorstore`: Binary file storing the vector database.
- `.gitignore`: Ignores virtual environment and cache files.
---
## Example Workflow
1. **Initialize the database:**
```sh
python main.py init
```
2. **Feed phrases:**
```sh
python main.py feed
```
Enter phrases interactively.
3. **Query for similar phrases:**
```sh
python main.py query
```
Enter a phrase to find similar ones.
4. **View corpus, vocabulary, and similarity matrix:**
```sh
python main.py view -c none -v none -m none
```
5. **Clear or delete the database:**
```sh
python main.py clear
python main.py delete
```
---
## Implementation Details
- **VectorStore Class:**
- Stores phrase vectors and similarity indices.
- Methods for adding vectors, updating similarity matrix, finding similar vectors, and viewing the similarity matrix.
- **DBOperations:**
- Handles reading/writing the vector store, feeding new phrases, querying, viewing, clearing, and deleting the database.
- **CLI:**
- Uses `argparse` for command parsing and interactive input for feeding/querying phrases.