DeepPavlov is an open-source library for building end-to-end dialogue systems and chatbots with deep learning. It bundles production-usable NLP components — named entity recognition, intent classification, question answering, slot filling and more — into trainable pipelines you can assemble into a working assistant. The project has about 7,000 stars on GitHub and was updated in 2025.
DeepPavlov grew out of the conversational AI research group at MIPT and is organized around reusable components: each NLP task ships as a configurable pipeline with pretrained models, so you can fine-tune a BERT-based intent classifier or NER model on your data with a config file rather than new code. For complete assistants, the companion DeepPavlov Agent framework orchestrates multiple skills — goal-oriented skills, chit-chat, question answering — with a dialogue manager that routes each user turn to the right skill and tracks state. Models can be served through REST APIs, and the library includes ready configurations for tasks in English and Russian, including the well-known SberQuAD-style reading comprehension setups. Everything is Apache-2.0 licensed and built on TensorFlow/PyTorch-era tooling, which makes it a solid, research-friendly foundation for teams that want full control over their conversational stack instead of a hosted chatbot black box.
DeepPavlov is completely free and open source under the Apache-2.0 license, including its pretrained models.
DeepPavlov is for NLP engineers and researchers building custom chatbots and dialogue systems who want proven, fine-tunable components rather than a closed hosted platform.