OpenAI Chatbot

This script is a chatbot that uses OpenAI's GPT-3 API to generate natural responses. It reads a text file containing chat messages from a Twitch channel, and generates a summary for the streamer.

Requirements

Yes. There are requirements. I'll add them here when I have a list. Obviously you'll need python installed. If you're on Linux, good luck. Audio doesn't work on WSL and I don't have an easy way to test yet. Post in issues and we'll try to figure it all out.

How to install

• Clone the repo
• Run install.bat or install.sh
• Setup a third party app to capture chat to a text file. (e.g. Streamdeck, TouchPortal, etc)
• Edit run.bat or run.sh to point to your chat log file and to supply your OpenAI key
• Launch run.bat or run.sh

How to Use

The script is run with the following parameters:

--chat_file: The path to the text file containing the chat messages from Twitch. Default is 'chat.txt'.

--streamer: The streamer's name (optional).

--streamer_twitch: The streamer's Twitch username (optional).

--streamer_pronouns: The streamer's pronouns (default: they/them).

--assistant: The assistant's name (optional).

--poll_interval: The interval to poll the chat file in seconds (default: 5).

--context: The context to be used for the OpenAI API request (default: "The bot is an assistant that will read the following stream chat and reply with a summary for the streamer.").

--context_file: The path to the text file containing the context to be used for the OpenAI API request (optional, if not provided the script will use the context argument).

--custom_pronunciations_file: The path to the JSON file containing custom pronunciations (optional).

--verbose: Print out the input text before sending it to OpenAI API (default: false).

--inactive_chat: How many seconds to wait for a chat message before the AI tries to engage chat. Set to 0 to disable (default: 30).

--command_users: The path to a text file containing user names that are allowed to use the !opentaai command. One user per line (optional).

--output_file: The path to a text file to be used by your 3rd party application to send text to the Twitch chat. One message per line. Defaults to "./output.txt".

--ignored_users_file: File to store ignored user names in (default: ignored_users.json).

--attitude: The current attitude for the bot. Defaults to neutral.

--timezone: Timezone you are in. List of timezones can be found at https://pythonhosted.org/pytz/#available-time-zones (default: US/Central).

--settings: The path to a json file with settings to load. If no settings file is specified, settings.txt will be created and updated with the settings at runtime.

--open_api_key: The OpenAI API key (optional, if not provided the script will use the api_key_file argument).

--api_key_file: The path to the file containing the OpenAI API key (optional, if not provided the script will use the open_api_key argument).

--model: The GPT-3 model to use (default: gpt-3.5-turbo).

--temperature: The setting for the model's temperature (default: 0.9).

--llm_api: Which LLM API to use (default: openai_api).

--llm_interval: Number of seconds between calls to the LLM. Defaults to 20.

--tts_rate: The speech rate in words per minute (default: 1.0).

--tts_volume: The speech volume from 0.0 to 1.0 (default: 1.0).

--tts_voice: The speech voice (default: 0).

--tts_model: The TTS model to use (default: tts_models/en/ljspeech/tacotron2-DDC_ph).

--coqui_studio_api_token: The path to a file containing your Coqui Studio API key (optional).

--tts_engine: Which TTS engine to use. Options: coqui, pyttsx3. Defaults to pyttsx3.

--force_tts_cpu: Force Coqui TTS to use CPU, even if CUDA is available. Ignore when using pyttsx3. Defaults to False.

--microphone: Select the microphone by index or name.

Example

python main.py --api_key_file api_key.txt --verbose --streamer Ubie --assistant "Stream AI" --context_file context.txt --custom_pronunciations_file custom_words.txt --inactive_chat 0 --coqui_studio_api_token coqui_api_key.txt --streamer_twitch theubie --command_users command_users.txt --tts_engine moegoe --tts_voice 384 --tts_rate 1.3 --llm_api openai_api --microphone "Microphone (NVIDIA Broadcast)" --attitude "snarky and sarcastic" --llm_interval 60

Future Plans

• Allow using text-generation-webui instead of OpenAI - Mostly Done!
• Allow using other local tts models - In progress. Coqui, Moegoe, and Pyttsx3 implemented, working on more.
• Integrate a chatbot directly to skip needing a 3rd party to genereate the chat log - Mostly Done!
• Money printer that goes brrr?