A simple standalone chat bot empowered by LLM services and AI models working with Nextcloud Talk App.

git clone https://github.com/CrazyShipOne/nextcloud_talk_pybot.git
cd nextcloud_talk_pybot
cp .env.example .env
pip install -r requirements.txt
./start.sh

Ensure edit .env by your credentials and tokens. refer to .env / Docker Environment Variable

git clone https://github.com/CrazyShipOne/nextcloud_talk_pybot.git
cd nextcloud_talk_pybot
docker build -t crazyshipone/nextcloud_talk_pybot .
docker run -d \
    -n nextcloud_talk_pybot \
    --restart=always \
    # Add environment variable by -e ENV_NAME=ENV_VALUE
    # -e NC_BASE_URL=https://www.mynextcloud.com:8080
    crazyshipone/nextcloud_talk_pybot

Ensure add environment variables refer to .env / Docker Environment Variable

NC_BASE_URL: Required. Base url of nextcloud endpoint. For example https://www.mynextcloud.com:8080.
NC_USERNAME: Required. User name of created bot account.
NC_PASSWORD: Required. Either user password or app password of created bot account.

POLL_INTERVAL: Optional. Message polling interval in seconds, default: 5.
ONLY_NEW: Optional. Set True to poll messages send after the bot is started, False to poll all unread messages, default: True.
MAX_MESSAGE: Optional. Maximum unread messages polled from one chat, default: 10

LOG_LEVEL: Optional. Output logging level, default: Info

MAX_CHAT_HISTORY: Optional. Maximum chat history stored, set 0 to not store. Caution: Set to a large number will cost lots of tokens! default: 0
HISTORY_STORAGE: Optional. Storage to save chat history, values are below. default: memory

memory: save in memory
redis: save in redis

REDIS_HOST: Required if HISTORY_STORAGE is 'redis'. Redis host.
REDIS_PORT: Optional. Redis port. default: 6379
REDIS_PASS: Optional. Redis password.
REDIS_DB: Optional. Redis db number. default: 0

Set if plugin will be used.

OPENAI_API_KEY: OpenAI's api key.

HFACE_TOEKN: HuggingFace's token.

1. Register a new account for chat bot in Nextcloud
2. Start the agent with new account's credentails
3. Open a new chat and invite chat bot
4. Type ![plugin]:[function] message to chat, for example !openai:chat Who are you?. Any message not start with ! will be responded with a help message.

• Add LLM plugins:
  • OpenAI (Partly finished)
  • Google Bard
  • Claude
  • Models on Huggingface
  • Models run on local machine
• Add Tool plugins:
  • Search result from google
  • Weather
• Add memory for LLM plugins