nndeploy is a C++ AI deployment framework with visualization, multi-end inference, and support for multiple backends. It offers Python/C++ APIs, a web UI, and JSON workflow export. Latest release updates and basic usage are documented in the repo.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
nndeploy is a simple-to-use, high-performance AI deployment framework that targets end-to-end deployment across desktop, mobile, edge devices, and servers. It emphasizes visual workflows, Python/C++ customization, and multi-end inference support, with a JSON workflow export feature and cross-platform deployment.
How it works
The project provides a visual workflow editor for deploying AI algorithms, with support for parallel execution patterns (serial, pipeline, task parallel) and memory optimizations (zero-copy, memory pool, reuse). It integrates multiple inference backends (ONNXRuntime, TensorRT, OpenVINO, MNN, TNN, ncnn, CoreML, AscendCL, RKNN, SNPE, TVM, PyTorch, plus internal modules) and allows selecting inference engines per workflow. Workflows can be exported to JSON and run via Python or C++ APIs.
Getting started
- Install:
pip install --upgrade nndeploy
- Start the visualization UI:
# 方式一：命令行
nndeploy-app --port 8000
# 方式二：代码启动
cd path/to/nndeploy
python app.py --port 8000
- Save/load workflows and run in production:
# Python CLI
nndeploy-run-json --json_file path/to/workflow.json
# C++ CLI
nndeploy_demo_run_json --json_file path/to/workflow.json
- Python/C++ example usage is provided for loading and running LLM workflows, including code snippets for loading a JSON workflow and executing it via a graph API.
Recent releases
- v3.0.10 (2026-04-04): 提升device模块稳定性
- v3.0.9 (2026-04-04): 完善GitHub CI
- v3.0.8 (2025-12-04): Fix the bug that arises from JSON deserialization
- v3.0.7 (2025-11-12): 前端/后端/底层架构优化及依赖升级
- v3.0.6 (2025-11-04): 强化构建/依赖和 OpenCV 升级
Traction
- Stars: 1856
Behind the repo
- Repository is maintained by nndeploy with multiple language bindings (Python, C++) and multi-platform support; mentions integration with numerous inference backends and a visual workflow ecosystem.
Caveats
- License: Apache-2.0
- Created: 2023-08-08
- Last push: 2026-08-03
- Open issues: 23
- Primary language: C++
- Requires Python 3.10+ and includes ONNXRuntime and MNN by default; other backends available in developer mode.