datapaintings/chatlog-marker

Web tool to mark chat dialogs by a set of topics / subtopics for later usage as a training set for neural network classifier

Used to mark chat dialogs by a set of topics / subtopics for later usage as a training set for neural network classifier. App itself starts with docker-compose and consists of django app itself (ran by gunicorn) and nginx as reverse proxy. It requires external PostgreSQL 10+ and Elasticsearch 5.6 + Kibana 5.6. Optionally you can add Consul 1.2.3 and Logstash 5.6.

Requirements:

docker django 2.0+ psycopg2 2.7.5

External services used:

1. PostgreSQL 10+
2. Consul 1.2.3 (Optional. Used only as KV, can be replaced by setting ENV)
3. Elasticsearch 5.6
4. Kibana 5.6
5. Logstash 5.6 with postgres plugin (Optional)

Запуск

1-a) Fill consul KV with settings

• DB_HOST=<postgres_host>
• DB_PORT=5432
• DB_NAME=bots
• DB_USER=
• DB_PASS=
• ELASTIC_URL=<host:port ELASTIC> - using this path collection dialogs will be created

fill app.env with:

• CONSUL_URL=<full URL to consul's http API, including folders>

if any of settings is not loaded from consul, app wil try to find all the settings in ENV

OR

1-b) Fill app.env with variables from 1-a

2. docker-compose up -d
3. docker exec dialog_marker_app python manage.py collectstatic