Third-party open source components used in CalfSystem Python scripts (plugins & local AI), with their license information. Software download and use are subject to these components as listed on this page.
The following open-source LLMs are available through the in-app model download feature. Model weight files are downloaded directly by the user's device from official open-source repositories (ModelScope / HuggingFace). CalfSystem does not own, distribute, or cache these model files.
+| Model | +Rights Holder | +License | +Commercial | +
|---|---|---|---|
| Qwen3 1.7B / 4B / 8B | Alibaba (Tongyi Qianwen) | Apache 2.0 | ✅ |
| Qwen2.5 7B Instruct | Alibaba (Tongyi Qianwen) | Apache 2.0 | ✅ |
| Mistral 7B Instruct v0.3 | Mistral AI | Apache 2.0 | ✅ |
| DeepSeek-R1-Distill-Qwen 7B | DeepSeek (深度求索) | MIT | ✅ |
| Phi-4 14B | Microsoft | MIT | ✅ |
| Gemma 2B / 7B Instruct | Gemma ToS | ✅ |
Full license notices are available in the software installation directory under Docs\ and Legal\ folders.
Local AI features and plugin scripts rely on the user's local Python environment. The software does not bundle Python or third-party Python packages; user-installed packages are subject to their respective licenses. The following third-party libraries may be used:
| Component | License | Purpose |
|---|---|---|
| torch | BSD-3-Clause | Deep learning framework (local model inference & training) |
| transformers | Apache-2.0 | Pretrained model loading & inference |
| peft | Apache-2.0 | LoRA and parameter-efficient fine-tuning |
| datasets | Apache-2.0 | Training dataset handling |
| bitsandbytes | MIT | Quantized training |
| accelerate | Apache-2.0 | Distributed training & inference |
| huggingface_hub | Apache-2.0 | Model download & caching |
| fastapi | MIT | Local AI API server |
| uvicorn | BSD-3-Clause | ASGI server |
| pydantic | MIT | Data validation |
| psutil | BSD-3-Clause | System resource monitoring |
| requests | Apache-2.0 | HTTP requests (API testing, plugins) |
| beautifulsoup4 | MIT | HTML parsing (plugin examples) |
Python interpreter is provided by the user's local installation; plugin scripts and documentation examples are project-developed or sample code.