maximophone / knowledgebot

KnowledgeBot

KnowledgeBot is a personal AI assistant that leverages Large Language Models (LLMs) to provide a text interface for interacting with LLMs and a system for transcribing and summarizing audio recordings.

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/KnowledgeBot.git
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Set up API keys:

   • Create a secrets.yml file in the project root directory.
   • Add the following keys to the file:

     claude_api_key: "your_claude_api_key"
     gemini_api_key: "your_gemini_api_key"
     openai_api_key: "your_openai_api_key"
     openai_org: "your_openai_organization"

   • Replace the placeholders with your actual API keys and organization.

Usage

Text Interface

To use the text interface for interacting with LLMs, run the following command:

python run_ai.py [--model MODEL_NAME]

• --model MODEL_NAME: (Optional) Specify the model to use. Default is "opus".

The script will monitor the conversation directory for changes. When a file is modified, it will process the conversation and generate a response using the specified model.

To use a different model, you can also prefix your text like so:

=#[model]

The available options are:

• opus
• sonnet
• haiku
• gpt4
• gpt3.5
• gemini1.0
• gemini1.5

Audio Transcription and Summarization

To transcribe and summarize audio recordings, run the following command:

python transcribe.py

The script will perform the following tasks:

1. Transcribe audio files from the Audio directory and save the transcriptions in the Transcriptions directory.
2. Improve the transcriptions and save the improved versions in the Improved directory.
3. Summarize the transcriptions and save the summaries in the Summaries directory.

The script will continuously monitor the directories and process new files as they are added.

Project Structure

• KnowledgeBot-main/: Project root directory
  • Audio/: Directory for storing audio files to be transcribed
  • Transcriptions/: Directory for storing transcriptions of audio files
  • Improved/: Directory for storing improved versions of transcriptions
  • Summaries/: Directory for storing summaries of transcriptions
  • conversation/: Directory for storing conversation files for the text interface
  • prompts/: Directory for storing prompt templates
  • ai.py: Module for interacting with different AI models
  • bot.py: Module for the KnowledgeBot functionality
  • config.py: Module for configuration management
  • repeater.py: Module for running repeated tasks
  • run_ai.py: Entry point for the text interface
  • summarisation.py: Module for summarizing transcriptions
  • transcribe.py: Entry point for audio transcription and summarization

License

This project is licensed under the MIT License.