inference-cli

What it is and what it does

Inference CLI is a lightweight command-line interface for deploying and running computer vision models locally or via Roboflow's hosted API. It wraps the parent inference package to provide server management (start, stop, status) and single-image inference commands without requiring custom Docker image building. The tool automatically detects your device (x86 CPU, ARM64, or NVIDIA GPU) and pulls the appropriate Docker image, handling the containerization details transparently.

You use it to start a local inference server on a specified port, then send images to it for predictions in JSON format. It supports object detection, classification, instance segmentation, and foundation models (CLIP, SAM), and can route requests to Roboflow's serverless hosted API instead. The CLI depends on 21 runtime packages including docker, click, typer for command parsing, opencv-python and pillow for image handling, and system introspection tools (nvidia-ml-py, py-cpuinfo) to select the right inference backend.

Use it for:

• Start a local inference server on your machine and run batch inference on images without writing Python code.
• Deploy YOLOv5 or YOLOv8 or custom Roboflow models to edge devices (ARM64, x86) using the CLI's automatic device detection.
• Run inference on a single image via command line and parse the JSON output for integration into shell scripts or CI/CD pipelines.
• Switch between local and hosted inference (Roboflow serverless) by changing the --host parameter without code changes.
• Monitor and manage inference server lifecycle (start, check status, stop) from the command line in production environments.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

A command-line tool for running computer vision inference locally via Docker or against Roboflow's hosted API, supporting object detection, classification, instance segmentation, and foundation models like CLIP and SAM.

Yes, if you need a straightforward CLI for deploying Roboflow-compatible vision models locally or to edge devices. Low install friction, active maintenance, and no known vulnerabilities make it reliable. However, verify that the GPL-3.0 and AGPL-3.0 licenses on bundled models (YOLOv5, YOLOv8) align with your project's licensing requirements before committing to production use.

Install

inference-cli on PyPI

pip install inference-cli

uv add inference-cli

poetry add inference-cli

Installing inference-cli

Before you install

Low install friction with a pure-Python wheel. Active maintenance as of 2026-08-14 with 2416 repository stars. Requires Python 3.10 or later and Docker for local server operation.

License in practice

Distributed under Apache 2.0, permissive and commercial-friendly. Note that individual models bundled with inference (YOLOv5, YOLOv8) carry AGPL-3.0 or GPL-3.0 licenses; verify compatibility with your use case.

Quickstart

pip install inference-cli

inference server start --port 9001
inference infer ./image.jpg --project-id my-project --model-version 1 --api-key my-api-key

Docker must be installed and running on your machine; Python 3.10 or later required.

Verify before relying

• Whether Docker must be pre-installed or if the CLI handles installation automatically.
• Whether the tool supports GPU inference without additional NVIDIA driver setup beyond nvidia-ml-py.
• Performance characteristics and latency for typical inference workloads on different device types.

Package facts

Evidence: inference_cli-1.4.1-py3-none-any.whl

Tags

More Software Development packages