DeepPavlov Dream

DeepPavlov Dream is a platform for creating multi-skill generative AI assistants.

To learn more about the platform and how to build AI assistants with it, please visit Dream. If you want to learn more about DeepPavlov Agent that powers Dream visit DeepPavlov Agent documentation.

Star History

Distributions

We've already included six distributions: four of them are based on lightweight Deepy socialbot, one is a full-sized Dream chatbot (based on Alexa Prize Challenge version) in English and a Dream chatbot in Russian.

Deepy Base

Base version of Lunar assistant. Deepy Base contains Spelling Preprocessing annotator, template-based Harvesters Maintenance Skill, and AIML-based open-domain Program-y Skill based on Dialog Flow Framework.

Deepy Advanced

Advanced version of Lunar assistant. Deepy Advanced contains Spelling Preprocessing, Sentence Segmentation, Entity Linking and Intent Catcher annotators, Harvesters Maintenance GoBot Skill for goal-oriented responses, and AIML-based open-domain Program-y Skill based on Dialog Flow Framework.

Deepy FAQ

FAQ version of Lunar assistant. Deepy FAQ contains Spelling Preprocessing annotator, template-based Frequently Asked Questions Skill, and AIML-based open-domain Program-y Skill based on Dialog Flow Framework.

Deepy GoBot

Goal-oriented version of Lunar assistant. Deepy GoBot Base contains Spelling Preprocessing annotator, Harvesters Maintenance GoBot Skill for goal-oriented responses, and AIML-based open-domain Program-y Skill based on Dialog Flow Framework.

Dream

Full version of DeepPavlov Dream Socialbot. This is almost the same version of the DREAM socialbot as at the end of Alexa Prize Challenge 4. Some API services are replaced with trainable models. Some services (e.g., News Annotator, Game Skill, Weather Skill) require private keys for underlying APIs, most of them can be obtained for free. If you want to use these services in local deployments, add your keys to the environmental variables (e.g., ./.env, ./.env_ru). This version of Dream Socialbot consumes a lot of resources because of its modular architecture and original goals (participation in Alexa Prize Challenge). We provide a demo of Dream Socialbot on our website.

Dream Mini

Mini version of DeepPavlov Dream Socialbot. This is a generative-based socialbot that uses English DialoGPT model to generate most of the responses. It also contains intent catcher and responder components to cover special user requests. Link to the distribution.

Dream Russian

Russian version of DeepPavlov Dream Socialbot. This is a generative-based socialbot that uses Russian DialoGPT by DeepPavlov to generate most of the responses. It also contains intent catcher and responder components to cover special user requests. Link to the distribution.

Prompted Dream Distributions

Mini version of DeepPavlov Dream Socialbot with the use of prompt-based generative models. This is a generative-based socialbot that uses large language models to generate most of the responses. You can upload your own prompts (json files) to common/prompts, add prompt names to PROMPTS_TO_CONSIDER (comma-separated), and the provided information will be used in LLM-powered reply generation as a prompt. Link to the distribution.

Quick Start

System Requirements

• Operating System: Ubuntu 18.04+, Windows 10+ (через WSL & WSL2), MacOS Big Sur;
• Version of docker from 20 and above;
• Version of docker-compose v1.29.2;
• Operative Memory from 2 Gb (using proxy), from 4 Gb (LLM-based prompted distributions) and from 20 Gb (old scripted distributions).

Clone the repo

git clone https://github.com/deeppavlov/dream.git

Install docker and docker-compose

If you get a "Permission denied" error running docker-compose, make sure to configure your docker user correctly.

Run one of the Dream distributions

Deepy Base

docker-compose -f docker-compose.yml -f assistant_dists/deepy_base/docker-compose.override.yml up --build

Deepy Advanced

docker-compose -f docker-compose.yml -f assistant_dists/deepy_adv/docker-compose.override.yml up --build

Deepy FAQ

docker-compose -f docker-compose.yml -f assistant_dists/deepy_faq/docker-compose.override.yml up --build

Deepy GoBot

docker-compose -f docker-compose.yml -f assistant_dists/deepy_gobot_base/docker-compose.override.yml up --build

Dream (via proxy)

The easiest way to try out Dream is to deploy it via proxy. All the requests will be redirected to DeepPavlov API, so you don't have to use any local resources. See proxy usage for details.

docker-compose -f docker-compose.yml -f assistant_dists/dream/docker-compose.override.yml -f assistant_dists/dream/dev.yml -f assistant_dists/dream/proxy.yml up --build

Dream (locally)

Please note, that DeepPavlov Dream components require a lot of resources. Refer to the components section to see estimated requirements.

docker-compose -f docker-compose.yml -f assistant_dists/dream/docker-compose.override.yml -f assistant_dists/dream/dev.yml up --build

We've also included a config with GPU allocations for multi-GPU environments:

AGENT_PORT=4242 docker-compose -f docker-compose.yml -f assistant_dists/dream/docker-compose.override.yml -f assistant_dists/dream/dev.yml -f assistant_dists/dream/test.yml up

When you need to restart particular docker container without re-building (make sure mapping in assistant_dists/dream/dev.yml is correct):

AGENT_PORT=4242 docker-compose -f docker-compose.yml -f assistant_dists/dream/docker-compose.override.yml -f assistant_dists/dream/dev.yml restart container-name

Prompted Dream

docker-compose -f docker-compose.yml -f assistant_dists/dream_persona_prompted/docker-compose.override.yml -f assistant_dists/dream_persona_prompted/dev.yml -f assistant_dists/dream_persona_prompted/proxy.yml up --build

We've also included a config with GPU allocations for multi-GPU environments.

Let's chat

DeepPavlov Agent provides several options for interaction: a command line interface, an HTTP API, and a Telegram bot

CLI

In a separate terminal tab run:

docker-compose exec agent python -m deeppavlov_agent.run agent.channel=cmd agent.pipeline_config=assistant_dists/dream/pipeline_conf.json

Enter your username and have a chat with Dream!

HTTP API

Once you've started the bot, DeepPavlov's Agent API will run on http://localhost:4242. You can learn about the API from the DeepPavlov Agent Docs.

A basic chat interface will be available at http://localhost:4242/chat.

Telegram Bot

Currently, Telegram bot is deployed instead of HTTP API. Edit agent command definition inside docker-compose.override.yml config:

agent:
  command: sh -c 'bin/wait && python -m deeppavlov_agent.run agent.channel=telegram agent.telegram_token=<TELEGRAM_BOT_TOKEN> agent.pipeline_config=assistant_dists/dream/pipeline_conf.json'

NOTE: treat your Telegram token as a secret and do not commit it to public repositories!

Configuration and proxy usage

Dream uses several docker-compose configuration files:

./docker-compose.yml is the core config which includes containers for DeepPavlov Agent and mongo database;

./assistant_dists/*/docker-compose.override.yml lists all components for the distribution;

./assistant_dists/dream/dev.yml includes volume bindings for easier Dream debugging;

./assistant_dists/dream/proxy.yml is a list of proxied containers.

If your deployment resources are limited, you can replace containers with their proxied copies hosted by DeepPavlov. To do this, override those container definitions inside proxy.yml, e.g.:

convers-evaluator-annotator:
  command: ["nginx", "-g", "daemon off;"]
  build:
    context: dp/proxy/
    dockerfile: Dockerfile
  environment:
    - PROXY_PASS=proxy.deeppavlov.ai:8004
    - SERVICE_PORT=8004

and include this config in your deployment command:

docker-compose -f docker-compose.yml -f assistant_dists/dream/docker-compose.override.yml -f assistant_dists/dream/dev.yml -f assistant_dists/dream/proxy.yml up --build

By default, proxy.yml contains all available proxy definitions.

Components English Version

Dream Architecture is presented in the following image:

Name	Requirements	Description
Rule Based Selector		Algorithm that selects list of skills to generate candidate responses to the current context based on topics, entities, emotions, toxicity, dialogue acts and dialogue history
Response Selector	50 MB RAM	Algorithm that selects a final responses among the given list of candidate responses

Annotators

Name	Requirements	Description
ASR	40 MB RAM	calculates overall ASR confidence for a given utterance and grades it as either very low, low, medium, or high (for Amazon markup)
Badlisted Words	150 MB RAM	detects words and phrases from the badlist
Combined Classification	1.5 GB RAM, 3.5 GB GPU	BERT-based model including topic classification, dialog acts classification, sentiment, toxicity, emotion, factoid classification
Combined Classification lightweight	1.6 GB RAM	The same model as Combined Classification, but takes 42% less time thanks to the lighter backbone
COMeT Atomic	2 GB RAM, 1.1 GB GPU	Commonsense prediction models COMeT Atomic
COMeT ConceptNet	2 GB RAM, 1.1 GB GPU	Commonsense prediction models COMeT ConceptNet
Convers Evaluator Annotator	1 GB RAM, 4.5 GB GPU	is trained on the Alexa Prize data from the previous competitions and predicts whether the candidate response is interesting, comprehensible, on-topic, engaging, or erroneous
Emotion Classification	2.5 GB RAM	emotion classification annotator
Entity Detection	1.5 GB RAM, 3.2 GB GPU	extracts entities and their types from utterances
Entity Linking	2.5 GB RAM, 1.3 GB GPU	finds Wikidata entity ids for the entities detected with Entity Detection
Entity Storer	220 MB RAM	a rule-based component, which stores entities from the user's and socialbot's utterances if opinion expression is detected with patterns or MIDAS Classifier and saves them along with the detected attitude to dialogue state
Fact Random	50 MB RAM	returns random facts for the given entity (for entities from user utterance)
Fact Retrieval	7.4 GB RAM, 1.2 GB GPU	extracts facts from Wikipedia and wikiHow
Intent Catcher	1.7 GB RAM, 2.4 GB GPU	classifies user utterances into a number of predefined intents which are trained on a set of phrases and regexps
KBQA	2 GB RAM, 1.4 GB GPU	answers user's factoid questions based on Wikidata KB
MIDAS Classification	1.1 GB RAM, 4.5 GB GPU	BERT-based model trained on a semantic classes subset of MIDAS dataset
MIDAS Predictor	30 MB RAM	BERT-based model trained on a semantic classes subset of MIDAS dataset
NER	2.2 GB RAM, 5 GB GPU	extracts person names, names of locations, organizations from uncased text
News API Annotator	80 MB RAM	extracts the latest news about entities or topics using the GNews API. DeepPavlov Dream deployments utilize our own API key.
Personality Catcher	30 MB RAM	the skill is to change the system's personality description via chatting interface, it works as a system command, the response is system-like message
Prompt Selector	50 MB RAM	Annotator utilizing Sentence Ranker to rank prompts and selecting N_SENTENCES_TO_RETURN most relevant prompts (based on questions provided in prompts)
Property Extraction	6.3 GiB RAM	extracts user attributes from utterances
Rake Keywords	40 MB RAM	extracts keywords from utterances with the help of RAKE algorithm
Relative Persona Extractor	50 MB RAM	Annotator utilizing Sentence Ranker to rank persona sentences and selecting N_SENTENCES_TO_RETURN the most relevant sentences
Sentrewrite	200 MB RAM	rewrites user's utterances by replacing pronouns with specific names that provide more useful information to downstream components
Sentseg	1 GB RAM	allows us to handle long and complex user's utterances by splitting them into sentences and recovering punctuation
Spacy Nounphrases	180 MB RAM	extracts nounphrases using Spacy and filters out generic ones
Speech Function Classifier	1.1 GB RAM, 4.5 GB GPU	a hierarchical algorithm based on several linear models and a rule-based approach for the prediction of speech functions described by Eggins and Slade
Speech Function Predictor	1.1 GB RAM, 4.5 GB GPU	yields probabilities of speech functions that can follow a speech function predicted by Speech Function Classifier
Spelling Preprocessing	50 MB RAM	pattern-based component to rewrite different colloquial expressions to a more formal style of conversation
Topic Recommendation	40 MB RAM	offers a topic for further conversation using the information about the discussed topics and user's preferences. Current version is based on Reddit personalities (see Dream Report for Alexa Prize 4).
Toxic Classification	3.5 GB RAM, 3 GB GPU	Toxic classification model from Transformers specified as PRETRAINED_MODEL_NAME_OR_PATH
User Persona Extractor	40 MB RAM	determines which age category the user belongs to based on some key words
Wiki Parser	100 MB RAM	extracts Wikidata triplets for the entities detected with Entity Linking
Wiki Facts	1.7 GB RAM	model that extracts related facts from Wikipedia and WikiHow pages

Services

Name	Requirements	Description
DialoGPT	1.2 GB RAM, 2.1 GB GPU	generative service based on Transformers generative model, the model is set in docker compose argument PRETRAINED_MODEL_NAME_OR_PATH (for example, microsoft/DialoGPT-small with 0.2-0.5 sec on GPU)
DialoGPT Persona-based	1.2 GB RAM, 2.1 GB GPU	generative service based on Transformers generative model, the model was pre-trained on the PersonaChat dataset to generate a response conditioned on a several sentences of the socialbot's persona
Image Captioning	4 GB RAM, 5.4 GB GPU	creates text representation of a received image
Infilling	1 GB RAM, 1.2 GB GPU	(turned off but the code is available) generative service based on Infilling model, for the given utterance returns utterance where _ from original text is replaced with generated tokens
Knowledge Grounding	2 GB RAM, 2.1 GB GPU	generative service based on BlenderBot architecture providing a response to the context taking into account an additional text paragraph
Masked LM	1.1 GB RAM, 1 GB GPU	(turned off but the code is available)
Seq2seq Persona-based	1.5 GB RAM, 1.5 GB GPU	generative service based on Transformers seq2seq model, the model was pre-trained on the PersonaChat dataset to generate a response conditioned on a several sentences of the socialbot's persona
Sentence Ranker	1.2 GB RAM, 2.1 GB GPU	ranking model given