This directory contains the backend services for the Blockly TFLite conversion feature.

• Python 3.8 or higher
• pip (Python package manager)

OR

• Docker and Docker Compose

1. Copy the environment file and configure:

cp .env.example .env
# Edit .env with your configuration

2. Build and run with Docker Compose:

docker-compose up -d

The service will be available at http://localhost:3000

To stop the service:

docker-compose down

To view logs:

docker-compose logs -f

1. Create a virtual environment (recommended):

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install dependencies:

pip install -r requirements.txt

3. Configure environment variables (create .env file):

cp .env.example .env
# Edit .env with your configuration

4. Run the application:

python app.py

Located in services/tflite_converter.py, this service provides functions to:

1. Convert TensorFlow.js models to TFLite format (convert_tfjs_to_tflite)

   • Accepts a TensorFlow.js model directory path
   • Converts to TensorFlow SavedModel (intermediate format)
   • Applies TFLite converter with optional quantization
   • Returns path to the generated .tflite file

2. Generate C/C++ byte arrays (tflite_to_c_array)

   • Reads a TFLite binary file
   • Generates formatted C/C++ code with const unsigned char array
   • Includes metadata comments and size constants
   • Returns formatted code string ready for microcontroller deployment

3. Extract model metadata (get_model_info)

   • Reads TFLite model and extracts input/output shapes
   • Returns dictionary with model information

from services.tflite_converter import convert_tfjs_to_tflite, tflite_to_c_array

# Convert TensorFlow.js model to TFLite
tflite_path = convert_tfjs_to_tflite(
    model_path='path/to/tfjs/model',
    output_path='output/model.tflite',
    quantize=True
)

# Generate C/C++ byte array
cpp_code = tflite_to_c_array(
    tflite_path=tflite_path,
    array_name='my_model_data',
    include_metadata=True
)

print(cpp_code)

Run tests using pytest:

pytest tests/

Run with coverage:

pytest --cov=services tests/

The conversion service will be exposed via a REST API endpoint (to be implemented in task 2.1):

• POST /api/convert-to-tflite
  • Accepts TensorFlow.js model data
  • Returns formatted C/C++ code or error response

• The conversion process requires temporary disk space for intermediate files
• Large models (>10MB) may take longer to convert
• Quantization is enabled by default to reduce model size for microcontrollers
• The service automatically cleans up temporary files after conversion