cmt-voice-za 1.0.0

African-language AI coding assistant for Python — converts spoken isiZulu, isiXhosa, Sesotho, Setswana, Sepedi, Tshivenda, Xitsonga to Python code

🎤 CMT_Voice

African-language AI coding assistant for Python and VS Code

"Funda iCSV ebizwa abafundi.csv bese ukhombisa imigqa emihlanu" → df = pd.read_csv("abafundi.csv") + print(df.head())

CMT_Voice lets South African students, researchers, and developers write Python by speaking isiZulu, isiXhosa, Sesotho, Setswana, Sepedi, Tshivenda, Xitsonga, and English — directly inside VS Code or Jupyter Notebooks.

---

✨ Features

• All Python keywords in 7+ South African languages
• All built-in functions (print, len, range, input, open, ...)
• pandas — read CSV/Excel, head, describe, groupby, merge, pivot ...
• NumPy — arrays, math, random, reshape, matrix ops ...
• Matplotlib + Seaborn — line, bar, histogram, scatter, pie, heatmap ...
• Scikit-Learn — train/test split, RandomForest, SVM, KNN, accuracy, PCA ...
• TensorFlow / Keras — Sequential, Dense, compile, fit, predict, save ...
• PyTorch — tensors, nn.Module, optimizers, backward, CUDA ...
• Flask — routes, GET/POST, jsonify, run ...
• File operations — open, read, write, os.path, makedirs ...
• AI fallback — Claude AI handles anything not in the dictionary
• VS Code extension — Ctrl+Shift+V to open the translation panel
• Multi-turn context — remembers your current file for smarter completions

---

🚀 Quick Start

Install Python package

pip install cmt-voice

Use in Python

import cmt_voice as cmtv

# Configure (reads ANTHROPIC_API_KEY env var automatically)
cmtv.configure()

# Translate simple phrase
print(cmtv.translate("Phrinta Sawubona"))
# → print("Sawubona")

# Translate with AI for complex intents
print(cmtv.translate("Funda iCSV ebizwa abafundi.csv bese ukhombisa imigqa emihlanu"))
# → df = pd.read_csv("abafundi.csv")
# → print(df.head())

# Interactive REPL
cmtv.listen_loop()

VS Code Extension

1. Copy the vscode_extension/ folder to your VS Code extensions directory
2. Set your API key in Settings → CMT_Voice → API Key
3. Open any .py file
4. Press Ctrl+Shift+V (Mac: Cmd+Shift+V) to open the panel
5. Type in any South African language → click Translate → Insert

---

🗣️ Language Examples

Keywords

Spoken	Language	Python
Phrinta Sawubona	isiZulu	print("Sawubona")
Khombisa igama	isiZulu	print(igama)
Laetsa lebitso	Sesotho	print(lebitso)
Bontsha leina	Setswana	print(leina)
Uma x inkulu kuno 5	isiZulu	if x > 5:
Xa x inkulu ku 5	isiXhosa	if x > 5:
Haeba x e fetisa 5	Sesotho	if x > 5:
Ngaphandle kwalokho	isiZulu	else:
Phinda i in range 10	isiZulu	for i in range(10):
Ngenkathi x inkulu kuno 0	isiZulu	while x > 0:
Dala umsebenzi obizwa sawubona	isiZulu	def sawubona():
Buyisa umphumela	isiZulu	return umphumela
iKilasi Mfundi	isiZulu	class Mfundi:
Zama	isiZulu	try:
Uma kunephutha Exception	isiZulu	except Exception:
Fakisa pandas	isiZulu	import pandas as pd
Iqiniso	isiZulu	True
Amanga	isiZulu	False
Lutho	isiZulu	None

Data Science (pandas)

Spoken	Python
Funda iCSV ebizwa data.csv	df = pd.read_csv("data.csv")
Funda iExcel ebizwa data.xlsx	df = pd.read_excel("data.xlsx")
Khombisa imigqa emihlanu yokuqala	print(df.head())
Bonisa ulwazi	df.info()
Chaza idatha	df.describe()
Ubunjalo bedatha	df.shape
Susa ulutho	df.dropna()
Hlela ngamanani	df.sort_values('column')
Hlanganisa ngemihlobo	df.groupby('column')
Gcina njengescsv	df.to_csv("output.csv")

Machine Learning (scikit-learn)

Spoken	Python
Hlukanisa idatha ibe yi-training ne-testing	X_train, X_test, y_train, y_test = train_test_split(...)
Dala i-random forest	model = RandomForestClassifier()
Dala i-logistic regression	model = LogisticRegression()
Dala i-svm	model = SVC()
Qeqesha imodeli	model.fit(X_train, y_train)
Bikezela	predictions = model.predict(X_test)
Bonisa ukunemba	print(accuracy_score(y_test, predictions))
Bonisa inhlalo	print(classification_report(...))
Lungisa idatha	scaler = StandardScaler()
Nciphisa izikhombisi	pca = PCA(n_components=2)

Visualisation (matplotlib/seaborn)

Spoken	Python
Dweba umugqa	plt.plot(x, y)
Dweba i-bar chart	plt.bar(x, y)
Dweba i-histogram	plt.hist(data)
Dweba amachashazi	plt.scatter(x, y)
Dweba i-pie chart	plt.pie(sizes)
Faka isihloko Imiphumela	plt.title("Imiphumela")
Dweba igraph	plt.show()
Dweba i-heatmap	sns.heatmap(df.corr())
Dweba i-pairplot	sns.pairplot(df)

Deep Learning (TensorFlow/PyTorch)

Spoken	Python
Dala i-neural network	model = tf.keras.Sequential()
Faka i-dense layer	model.add(tf.keras.layers.Dense(64, activation='relu'))
Faka i-dropout	model.add(tf.keras.layers.Dropout(0.5))
Lungisa imodeli	model.compile(optimizer='adam', ...)
Qeqesha imodeli	model.fit(X_train, y_train, epochs=10)
Hlola imodeli	loss, acc = model.evaluate(X_test, y_test)
Gcina imodeli	model.save("imodeli.h5")

---

🏗️ Architecture

CMT_Voice/
├── cmt_voice/
│   ├── __init__.py          # Public API
│   ├── engine.py            # Core engine (local + AI translation)
│   └── mappings.py          # 1000+ language → Python mappings
├── vscode_extension/
│   ├── package.json         # Extension manifest
│   └── src/extension.js     # VS Code extension (Ctrl+Shift+V)
├── examples/
│   ├── all_examples.py      # Full test suite
│   └── data_science_izulu.py # Complete isiZulu ML workflow
├── setup.py
└── README.md

Translation pipeline:

Spoken SA Language
       ↓
  LocalTranslator (dictionary, instant)
       ↓ (no match)
  AITranslator (Claude API, handles complex/novel phrases)
       ↓
  Post-processing (clean indentation, strip fences)
       ↓
  Valid Python Code → inserted at cursor in VS Code

---

⚙️ Configuration

import cmt_voice as cmtv

cmtv.configure(
    api_key="sk-ant-...",        # or set ANTHROPIC_API_KEY env var
    model="claude-sonnet-4-6",   # Claude model
    always_use_ai=False,         # True = always use AI (slower, more accurate)
)

VS Code settings (settings.json):

{
  "cmtvoice.apiKey": "sk-ant-...",
  "cmtvoice.model": "claude-sonnet-4-6",
  "cmtvoice.alwaysUseAI": false,
  "cmtvoice.addImportsAutomatically": true,
  "cmtvoice.insertAtCursor": true,
  "cmtvoice.language": "auto"
}

---

🌍 Supported Languages

Language	Region
isiZulu	KwaZulu-Natal, Gauteng
isiXhosa	Eastern Cape, Western Cape
Sesotho	Free State, Lesotho
Setswana	North West, Botswana
Sepedi (Northern Sotho)	Limpopo
Tshivenda	Limpopo
Xitsonga	Limpopo, Mpumalanga
English	All provinces
Mixed SA	Any combination

---

🎯 Roadmap

• v1.0 — Python package with 1000+ mappings
• v1.0 — VS Code extension (Ctrl+Shift+V)
• v1.1 — Jupyter Notebook extension (speak → cell code)
• v1.2 — Real microphone input (speech-to-text → code)
• v1.3 — More languages (Siswati, isiNdebele, Afrikaans)
• v2.0 — Full IDE plugin (JetBrains, Cursor)
• v2.0 — Voice activation ("Hey CMT") support

---

📄 License

MIT License — built for African students, researchers, and developers.

Sala kahle. Totsiens. Goodbye.