Repository navigation
v0.1.0 — NeuroServe: First Public Preview
Release date: 2025-09-13
NeuroServe is now ready as a reusable FastAPI AI inference server core.
This release focuses on a clean architecture, a modular plugin system, and cross-platform readiness (GPU/CPU/MPS/ROCm).
Highlights
- Clean architecture: core config/logging/errors, runtime device layer, unified responses.
- Plugin system: build & load model/services as plugins; ships with example
dummyplugin. - Cross-platform: CUDA (Linux/Windows), CPU everywhere, MPS on macOS.
- CI: Ubuntu, Windows, macOS + self-hosted GPU lane.
- Docs: Updated README with badges table, API guide, and full architecture doc.
Install & Run
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt
uvicorn app.main:app --reload --host 0.0.0.0 --port 8000