Discord Bot for Real-Time Hate Speech Detection using Ollama

Discord bot for user monitoring and management with Ollama-based, LLM-agnostic hate speech detection.

Installation and environment

0. [OPTIONAL] Run the Ollama client in the Docker hosting machine.

1. Assign the "NODE_ENV" environment variable as following export NODE_ENV="development/production/local" with the following format:

   NODE_ENV=
   DISCORD_OAUTH2_TOKEN=
   DISCORD_CLIENT_ID=
   CASSIE_HOST=
   CASSIE_PORT=
   CASSIE_KEYSPACE=
   OLLAMA_API_HOST=
   OLLAMA_API_PORT=
   OLLAMA_API_MODEL=
   OLLAMA_API_MODEL_SESSION_ID=
   COMPOSE_PROJECT_NAME=# optional
   CLIENT_LOCALE=# optional

   1.1 ENV_FILE values respectively by NODE_ENV value given:

   NODE_ENV	ENV_FILE
   production	.env
   development	.env.dev
   test	.env.dev
   "any"	.env.local

   1.2 Assign ANALYSIS_MODEL environment variable to be pulled with the Ollama API service container initialization. If not it will be set as default that's "gemma3:1b".

2. Run docker compose up at root directory.

   2.1 Optionally assign the ANALYZE_MESSAGE_SUCCESS_RATIO environment variable with a decimal number between 0 and 1 to adjust the analysis success ratio (default is 0.95 = 95%). IMPORTANT: this will determinate the application docker compose up outcome and depends entirely and directly on both the ratio and model used respectively.