StreamerAI

StreamerAI is a Python-based application that allows users to engage with an AI-powered streaming assistant. The application can fetch product scripts, answer comments, and handle text-to-speech (TTS) functionalities. It utilizes SQLite for database management, GPT-based models for answering questions, and TTS services for speech synthesis.

Overview

The main components of the StreamerAI application include:

• StreamerAI class: Represents the main functionality of the streamer AI, handling script fetching, comment answering, and TTS management.
• StreamCommentsDB class: A database utility class for managing stream comments using SQLite.
• TextToSpeech class: A TTS service wrapper that handles speech synthesis for the AI-generated responses.
• Chains class: A GPT-based question-answering module that creates context-aware AI responses.
• Retrieval class: A GPT-based question-answering module that retrieves relevant information based on user queries.

Setup

1. Set OpenAI API Key environment variable via export OPENAI_API_KEY=<key here>
2. Set Pinecone API Key environment variable via export PINECONE_API_KEY=<key here>
3. Set TTS API Key environment variable via export TTS_ACCESS_TOKEN=<key here>
4. Install poetry if you have not yet via pip install poetry
5. poetry env use python3.10
6. poetry shell to activate virtual environment
7. poetry install to install dependencies

Streamer Usage

1. poetry shell if you haven't yet
2. source .env if you haven't yet
3. Bootstrap database with some product data
   1. poetry run bootstrap --reset
   2. poetry run bootstrap --products
   3. poetry run bootstrap --personas
4. Start main script poetry run start --room_id ROOM_ID [--voice_type VOICE_TYPE] [--voice_style VOICE_STYLE] --platform {douyin,bili,fake} [--disable_script] [--persona PERSONA]

Testing without actually streaming on a platform

1. Start main script `poetry run start --room_id 'room id here'
2. poetry run fake --room_id 'room_id_here'
3. Type in some comments and see the AI respond