Zamia AI

Free and open source A.I. system based on Python, TensorFlow and Prolog.


The Zamia AI project is a framework that provides a set of components needed to build free, open source end-to-end speech and natural language processing A.I. systems.

Cloudless AI

Zamia AI enables you to build entirely self-contained A.I. systems with no dependencies on proprietary components, data sources or cloud services that you do not control.

Of course, you can always add or replace individual components with proprietary or cloud based ones if you wish. Conversely, you can take it to the other extreme and build completely self-contained systems that can run entirely offline.

Use Cases

Zamia AI can be applied to a wide variety of domains ranging from speech recognition to deep text analysis such as:

  • Chatbots
  • Virtual assistants
  • Question answering
  • Information extraction
  • Summarization
  • Classification
  • Sentiment analysis

Advanced features include deep natural language understanding, knowledge based reasoning, speech recognition, speech synthesis and more.

Ready to go

Zamia AI comes with the tools needed to create, train, test and maintain the neural network natural language models for the system.

Also included are Kaldi ASR (speech recognition) models for English and German.

Free and Open Source

All components and provided models are free - the code is open source under the Apache 2.0 license, all training data used to build our models is open.

All models included in Zamia AI are free (including open source training data) and cover many application domains.

View details »



Zamia AI's modular architecture consists of a stack of proven solutions to common AI tasks:

  • Powerful tools are provided to generate training samples from structured data such as RDF/LDF sources (e.g. WikiData), AIML and other corpora.
  • Natural language processing is performed by state of the art deep neural networks.
  • For reasoning an innovative combination of imperative Python code and declarative XSB Prolog is used.
  • Knowledge bases consist of prolog code - which for the most part can be generated from structured data.
  • For ASR (speech recognition, STT) state of the art Kaldi ASR nnet3 chain models are used. For speech output you have a choice of proven, free speech synthesizers (eSpeak NG, MaryTTS, SVOX Pico).

View details »

Get started

You can start exploring and using Zamia AI right now!

Look at our installation guide to get it up and running on your machine:

Getting Started »