Acknowledgments. This project would not have been possible without the decisive contribution of the free and open-source software listed below — foundational, shared by their authors with everyone. We express our deepest gratitude to them. We also thank Agent Zero, the intelligent assistant that supported the design and development of this project.
The product's own code remains deliberately closed: we have chosen to keep control of its evolution in the face of possible undesired uses. The dependencies themselves remain open source and compliant with their respective licenses.
Digital Partner relies on an ecosystem of open-source components. This page lists dependencies, their respective licenses, and how they are used in the project. It serves as a compliance register for license audits.
1. Python Dependencies — System Core
Component
License
Use in Tosio
Role
NumPy
BSD-3
Vector math, W topology, HNSW
Library
Flask
BSD-3
REST API backend
Framework
PyYAML
MIT
YAML governance DSL
Library
pytest
MIT
Test framework
Tool
LiteLLM
MIT
Multi-provider LLM abstraction
Library
scikit-learn
BSD-3
PCA, calibration, metrics
Library
FastAPI
MIT
Additional API endpoints
Framework
Cython
Apache-2.0
Native compilation of critical modules
Tool
Transformers / BERT(provided by HuggingFace)
Apache-2.0
Semantic embeddings, classification
Model
Pydantic
MIT
Data validation
Library
httpx
BSD-3
Async HTTP client
Library
aiohttp
Apache-2.0
Async HTTP client/server
Library
SQLAlchemy
MIT
Database ORM
Library
Uvicorn
BSD-3
ASGI server for the API
Server
python-jose
MIT
JOSE/JWT (multiple algorithms)
Library
webauthn
BSD-3
WebAuthn / Passkeys authentication
Library
scipy
BSD-3
Scientific computing, optimization
Library
hnswlib
Apache-2.0
Native HNSW vector index
Library
faiss-cpu
MIT
FAISS vector index fallback
Library
1bis. AI & Heavy Compute (ML ecosystem)
Component
License
Use in Tosio
Role
PyTorch
BSD-3
Deep learning, embeddings, models
Library
torchvision
BSD-3
Vision, image transformations
Library
Accelerate
Apache-2.0
Multi-device orchestration
Library
Diffusers
Apache-2.0
Diffusion generative models
Library
safetensors
Apache-2.0
Secure weight serialization
Library
open_clip_torch
MIT
Local CLIP image/text encoding
Model
soundfile
BSD-3
Audio read/write
Library
pydub
MIT
Audio processing
Library
pyzbar
MIT
QR / barcode decoding
Library
PyTorch Geometric
MIT
GNN (GATv2) for optimization guidance
Library
XGBoost
Apache-2.0
Tour cost regressor (meta-learning) — research only, not in production