Skip to content

v0.1.0 – NeuroServe: First Public Preview

Latest

Choose a tag to compare

@TamerOnLine TamerOnLine released this 13 Sep 20:28
· 6 commits to main since this release

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 dummy plugin.
  • 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