MAIA

MAIA (Memory-Augmented Intelligent Assistant) implements prompt chaining with enhanced long-term recall capabilities for LLM-powered intelligent assistants.

This work has been accepted to IUI 2025 (ACM Conference on Intelligent User Interfaces). 📄 Read the paper. Citation details are available in the Citation section.

Requirements

• Python 3.9
• CUDA 11.7
• PyTorch
• Transformers
• Additional dependencies listed in environment.yml or requirements.txt

Installation

Install dependencies using either conda:

conda env create --file environment.yml

Or pip:

pip install -r requirements.txt

Configuration

Create a .env file with the following environment variables:

SSL_CERT_PATH={Path to SSL fullchain}
SSL_KEY_PATH={Path to SSL privkey}
OPENAI_API_KEY={OpenAI API Key}
PALM_API_KEY={PaLM API Key}
GOOGLE_TTS_API_KEY={Google Cloud Text-to-Speech API Key}

Implementation

MAIA's core framework is implemented in the MAIAPrompter class (conversation/prompter.py). Key features include:

• Context extraction using pretrained transformer attention
• Short-term and long-term memory management
• Multi-step reasoning with dynamic prompt generation
• DPR-based memory retrieval

Usage

Web Interface

Launch the Gradio-based web interface for voice and text interactions:

python run_gradio.py [--server_name SERVER_NAME] [--server_port PORT] [--share]

Optional arguments:

• --server_name: Specify server name (default: 0.0.0.0)
• --server_port: Specify port number
• --share: Enable Gradio's share feature

Console Interface

Start text-based interactions in console:

python app.py

Architecture

MAIA consists of four main components:

1. Context Extraction: Pretrained transformer attention mechanism
2. Memory Module: DPR-based retrieval with STM/LTM
3. Prompt Generation: Dynamic multi-step reasoning chain
4. Response Generation: Template-based response synthesis

Citation

@inproceedings{seo2025prompt,
  author = {Seo, Seongbum and Yoo, Sangbong and Jang, Yun},
  title = {A Prompt Chaining Framework for Long-Term Recall in LLM-Powered Intelligent Assistant},
  year = {2025},
  month = mar,
  isbn = {9798400713064},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3708359.3712117},
  doi = {10.1145/3708359.3712117},
  booktitle = {Proceedings of the 30th International Conference on Intelligent User Interfaces},
  pages = {89–105},
  numpages = {17},
  keywords = {Intelligent Assistants, Large Language Models, Prompt Chaining},
  location = {Cagliari, Italy},
  series = {IUI '25}
}

License

MIT License - see LICENSE file for details