| # Edge Conversion Scripts | |
| This directory contains optional utility scripts to convert the **Orqanta AI / Psycho** model to formats like TFLite or ONNX if required. | |
| > **Note**: Orqanta AI already features precompiled native C/C++ engine binaries in `bin/android-arm64/orqanta` and `bin/linux-x86_64/orqanta` that run natively on Android (Car / Smartwatch) and Linux laptops with **<80MB RAM and sub-50ms CPU latency**, without needing TFLite or ONNX! | |
| ## Scripts | |
| | Script | Target Platform | Description | | |
| |--------|----------------|-------------| | |
| | `convert_to_tflite.py` | Android‑based car (or any Android device) | Optional exporter bridge to TensorFlow Lite (`.tflite`). | | |
| | `convert_to_onnx.py` | Linux laptop (or any Linux system) | Optional exporter bridge to ONNX (`.onnx`). | | |
| ## Requirements | |
| - **Python 3.9+** | |
| - For TFLite conversion: `pip install tensorflow` | |
| - For ONNX conversion: `pip install torch onnx` | |
| ## Direct Native Edge Execution (Recommended) | |
| Instead of converting, use the zero-overhead native engine directly: | |
| ```bash | |
| # On Linux Laptops (x86_64): | |
| ./bin/linux-x86_64/orqanta --model models/orqanta.cact --tools examples/linux_laptop/laptop_tools.json --prompt "Check battery" | |
| # On Android Automotive / Smartwatch (ARM64): | |
| adb push bin/android-arm64/orqanta /data/local/tmp/ | |
| adb push models/orqanta.cact /data/local/tmp/ | |
| adb shell "/data/local/tmp/orqanta --model /data/local/tmp/orqanta.cact --prompt 'Hello'" | |
| ``` | |