telekinesis-ai 0.0.1rc1

Official Telekinesis SDK for working with Telekinesis APIs.

Telekinesis Examples | Telekinesis Data

Telekinesis SDK

Telekinesis SDK is a modular Python-based SDK for Physical AI, providing a unified set of algorithms for robotics, 3D perception, computer vision, motion planning, and vision-language models.

It is designed for roboticists and computer vision engineers who want to build end-to-end Physical AI systems without stitching together fragmented libraries.

What You Can Build With Telekinesis SDK

Telekinesis SDK includes:

• 3D perception (filtering, registration, clustering)
• 2D perception (image processing, segmentation)
• Synthetic data generation
• Motion planning, kinematics, and control
• Vision-Language Models (VLMs)
• Physical AI agents

Learn more about the Telekinesis SDK in the About Telekinesis.

Installation

You will need to have Python 3.11 or higher set up to use the Telekinesis SDK

Run the following command to install the Telekinesis SDK:

pip install telekinesis-ai

Getting Started

Telekinesis SDK requires a valid API key to authenticate requests.

If you have not yet set up your API key, follow the official Quickstart Guide.

The recommended way to explore Telekinesis SDK today is via the telekinesis-examples repository, which contains fully runnable workflows built on top of the SDK.

Minimal Usage Example

The following example assumes the API key has been generated and has been set as TELEKINESIS_API_KEY environment variable.

Run a python code to quickly test your installation:

This example will fail if TELEKINESIS_API_KEY is not set correctly.

import numpy as np
from telekinesis import vitreous

# Create a cylinder mesh
cylinder_mesh = vitreous.create_cylinder_mesh(
		radius=0.01,
		height=0.02,
		radial_resolution=20,
		height_resolution=4,
		retain_base=False,
		vertex_tolerance=1e-6,
		transformation_matrix=np.eye(4, dtype=np.float32),
		compute_vertex_normals=True,
	)

# Convert it to point cloud
point_cloud = vitreous.convert_mesh_to_point_cloud(
		mesh=cylinder_mesh,
		num_points=10000,
		sampling_method="poisson_disk",
		initial_sampling_factor=5,
		initial_point_cloud=None,
		use_triangle_normal=False,
	)
print(point_cloud.positions)
# Use point_cloud in downstream processing or visualize the point cloud with any tool        

Expected output:

2025-12-22 12:45:36.914 | WARNING  | datatypes.serializer:serialize_to_pyarrow_ipc:20 - Attribute 'initial_point_cloud' is None, skipping serialization.
[[-3.8715214e-03  9.1124466e-03  3.9994060e-03]
 [ 9.1133742e-03 -3.8697000e-03  4.5292359e-03]
 [-4.8016016e-03 -8.6385468e-03  2.1635974e-03]
 ...
 [-2.8871899e-04  9.9542709e-03  2.2834308e-04]
 [-7.8424560e-03  6.1255670e-03  6.7548160e-03]
 [ 8.2263970e-05 -9.9869706e-03  5.0877198e-03]]

You are now set up to build with Telekinesis.

Resources

You can find all the information on Telekinesis SDK and the usage on Telekinesis Documentation.

We provide a large set of examples to help you learn all the available skills in the SDK. Explore other example scripts demonstrating Telekinesis SDK usage on our GitHub repositories - https://github.com/telekinesis-ai/telekinesis-examples.git.

To access the sample data visit our https://github.com/telekinesis-ai/telekinesis-data.

Support

For issues and questions:

• Create an issue in the GitHub repository Telekinesis Examples.
• Contact the Telekinesis development team.

GitHub  •  LinkedIn  •  X  •  Discord