Chat99 AI Assistant

Chat99 is an intelligent AI assistant with advanced memory, multi-model capabilities, and dynamic routing inspired by RouteLLM. My project has now diverged too much to practicilly integrate RouteLLM but check them out https://github.com/lm-sys/RouteLLM (I don't know these guys but just a shoutout).

Agent99

Chat99 is a pre-curser to the intended end goal of creating Agent99 and an agentic framework that dynamically routes between local/lower-end models for simpler or specific task execution while also assigning model parameters to achieve the best results for each specific task.

Agent99, as a project name, is a loose call back to the 60's comedy show "Get Smart". Get Smart was and probably still is on re-runs every afternoon, and I'd watch it after school. And like many of you... Agent 99 was of particular interest. Although that wasn't part of why I named it that. It was the first thing that popped into my head when I thought, "Agent... Agent... what?"

Scroll to Chat99 Readme, Agent99 still under Development

Key Features

• Dynamic model selection using RouteLLM
• Advanced memory management for context retention
• Support for multiple AI models (Claude, GPT-4, etc.)
• Calibration tool for optimizing model selection thresholds
• Local model support using Ollama

Dev notes

• memory_manager.py is a temporary solution so that focus can be placed on other project areas.

Installation Steps

1. Clone the repository:

   git clone https://github.com/GaryOcean428/Agent99.git
   cd Agent99

2. Install the required dependencies:

   pip install -r requirements.txt

3. Set up your environment variables by copying the .env.template file to .env and filling in your API keys:

   cp .env.template .env

   Then edit the .env file with your actual API keys.

How to Use

1. Run the calibration script to find the optimal threshold for model selection:

   python calibrate_threshold.py --sample-queries sample_queries.json --router mf --strong-model-pct 0.5

2. Start the Chat99 assistant:

   python main.py --use-dynamic-routing --router mf --threshold <threshold_from_calibration>

Running with Docker

To run the project using Docker, follow these steps:

1. Make sure you have Docker and Docker Compose installed on your system.

2. Build and run the Docker container:

   docker-compose up --build

   This command will build the Docker image and start the container.

3. Access the application by opening a web browser and navigating to http://localhost:5000.

4. To stop the container, press Ctrl+C in the terminal where docker-compose is running, or run:

   docker-compose down

How to Contribute

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Chat99

Chat99 is an intelligent AI assistant with advanced memory, multi-model capabilities, and dynamic routing using AdvancedRouter.

Features

• Dynamic model selection using AdvancedRouter
• Advanced memory management for context retention
• Support for multiple AI models (Claude, Llama, etc.)
• Calibration tool for optimizing model selection thresholds
• Local model support using Ollama

Installation

1. Clone the repository:

   git clone https://github.com/GaryOcean428/Agent99.git
   cd Agent99

2. Install the required dependencies:

   pip install -r requirements.txt

3. Set up your environment variables by copying the .env.template file to .env and filling in your API keys:

   cp .env.template .env

   Then edit the .env file with your actual API keys.

Usage

1. Run the calibration script to find the optimal threshold for model selection:

   python calibrate_threshold.py --sample-queries sample_queries.json --strong-model-pct 0.5

2. Start the Chat99 assistant:

   python chat99.py

3. Interact with the assistant by typing your messages when prompted.

4. To exit the program, you can:

   • Type "exit" when prompted for input
   • Use the keyboard interrupt (Ctrl + C on most systems, Cmd + C on macOS)

Project Structure

• chat99.py: Main script for the chat interface
• advanced_router.py: Handles dynamic model selection
• config.py: Configuration settings for the project
• memory_manager.py: Manages conversation context and memory
• models.py: Defines available AI models
• calibrate_threshold.py: Script for calibrating the routing threshold

Current Status

• Functional chat interface with dynamic model routing
• Support for multiple AI models (Claude 3.5, Llama 3.1 70B, Llama 3.1 8B)
• Basic complexity assessment for query routing
• Calibration script for optimizing routing thresholds

Future Development

• Enhance complexity assessment with more sophisticated NLP techniques
• Implement dynamic allocation of temperature and top-p parameters
• Develop multi-dimensional routing considering both model and parameter selection
• Integrate user preferences for response style
• Implement adaptive learning for continuous improvement of routing decisions
• Explore multi-model ensemble responses for comprehensive answers

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License Information

This project is licensed under the MIT License - see the LICENSE file for details.