synapseTools 0.1.32

Utility toolkit for data exploration, audio mel-spectrogram generation, and Spanish phonetic processing.

synapseTools

Data exploration, audio features, and Spanish phonetics in one toolbox

---

Overview

synapseTools is a utility library maintained by SYNAPSE AI SAS that currently includes:

• Exploratory data analysis (EDA): functions for null detection, outlier analysis, correlation heatmaps, and PCA visualization.
• Audio processing: mel-spectrogram generation and visualization for audio workflows.
• Spanish phonetics: phoneme and accent transformations tailored for Rioplatense Spanish.

This is an evolving project designed as the foundation for a comprehensive framework to work with data, AI, and autonomous agents. We are continuously expanding the toolkit with new features and capabilities.

Core principles:

• Practical – simple, focused functions with sensible defaults.
• Composable – integrate easily into your pipelines and notebooks.

---

Installation

pip install synapseTools

The package targets Python 3.8+.

If you want to install only specific feature sets you can use extras:

# Full installation (equivalent to base install)
pip install "synapseTools[all]"

# Only EDA utilities
pip install "synapseTools[eda]"

# Only audio / mel-spectrogram utilities
pip install "synapseTools[mel]"

# Only phoneme / accent tools
pip install "synapseTools[phonemes]"

For reproducible environments, we recommend using python -m venv or a tool like uv, poetry, or pipenv.

---

Modules

synapse_tools.eda

Utilities for quick exploratory data analysis on pandas.DataFrame objects.

• nulls(data, column) – prints count and percentage of nulls in a column.
• outliers(data, column, ...) – histogram + boxplot + descriptive statistics and IQR-based outlier detection, with optional dictionary output.
• heatmap_correlation(data, columns, ...) – Spearman/Pearson correlation heatmap with save/show options.
• pca_view(data, dimensions, target=None, ...) – runs PCA (2D or 3D) with optional scaling and target coloring.

These functions are handy when you want fast, visual feedback about a dataset without writing a lot of boilerplate plotting code.

synapse_tools.mel_spectrograms

Helpers for turning audio files into mel spectrograms and plotting them.

• load_audio_to_mel(file_path, sr=22050, ...) – loads an audio file, normalizes it, and returns a mel spectrogram as a NumPy array.
• graph_mel_spectrogram(spectrogram, output_dir='', name='Spectrogram', ...) – visualizes (and optionally saves) a mel spectrogram image.

This is especially useful in speech and TTS workflows where you need a repeatable way to extract and inspect mel features.

synapse_tools.phonemes

Rule-based utilities focused on Rioplatense Spanish (Argentina / Uruguay).

• phoneme(text, punctuation=False) – converts Spanish text into a simplified phoneme sequence using deterministic rules.
• accent(text, punctuation=False) – applies prosodic accentuation based on Spanish stress rules.
• dictionaries(text, order_by_frequency=True, pad=True) – builds phoneme-to-index and frequency dictionaries for modeling.
• phoneme_graphs(tokens, quantity, ...) – bar chart of phoneme frequencies.
• embeddings(input_dim, output_dim, std, ...) – frequency-aware initialization matrix for embedding layers.

These tools are designed to play nicely with downstream NLP / TTS models, where you often need custom tokenization and embeddings.

---

Basic usage

import pandas as pd
from synapse_tools import nulls, outliers, heatmap_correlation, load_audio_to_mel, graph_mel_spectrogram, phoneme

# EDA
df = pd.read_csv("data.csv")
nulls(df, "age")
outliers(df, "salary")

# Audio
mel = load_audio_to_mel("audio.wav")
graph_mel_spectrogram(mel, name="example")

# Phonetics (Rioplatense Spanish)
phoneme_text = phoneme("Esta es una oración de prueba.")
print(phoneme_text)

---

Contributing

If you want to contribute, see CONTRIBUTING.md.

---

License

This project is licensed under the Apache 2.0 license. See LICENSE for details.