Archived. This repository contains the demo code and slides from Scottish Summit 2024. It has been superseded and will be replaced with a new repository for Scottish Summit 2026 in Edinburgh, updated with multi-agent patterns and AI gateway architecture.
See SecuringTheRealm/str-agentic-adventures for the current active project.
The slides from Scottish Summit 2024 can be found here Link to slides.
Dungeons & Dragons (D&D) has long thrived on the creativity and guidance of the Dungeon Master (DM), orchestrating adventures for players. Recent advancements in AI and large language models (LLMs) present a new opportunity: a DM-less D&D experience. This session focuses on the practical application of LLMs and SLMs to facilitate an autonomous, engaging tabletop RPG experience, allowing every player to take in the narrative.
We'll explore the technical architecture behind deploying LLMs and fine tuned SLMs for DM-less gameplay, emphasising the integration of guardrails and memory systems for coherent storytelling. Key to this is maintaining a safe and immersive environment; so we'll also cover security measures such as red teaming and jailbreak prevention techniques that safeguard against exploits and ensure content appropriateness.
The technical overview will include audio inputs for voice and natural language understanding, enhancing player-AI interaction. Content generation by language models will be examined, demonstrating how AI can dynamically adapt stories, environments, and challenges in response to player actions.
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt- DEMO1: Link to ChatGPT
- DEMO2 - 3: Azure OpenAI + Basic chat interface; plus security features
- DEMO4: Azure OpenAI + Image generation
- DEMO5: Azure OpenAI + Image generation + Voice
- DEMO6: Azure OpenAI + Image generation + Voice + SLMs
- DEMO7: Azure OpenAI + Image generation + Voice + SLMs + RAG v Finetuning
- DEMO8: Azure OpenAI + Image generation + Voice + SLMs + RAG v Finetuning + Memory
- 3D https://openai.com/index/point-e/
- Adaptive cards - https://learn.microsoft.com/en-us/adaptive-cards/?WT.mc_id=AI-MVP-5004204 https://adaptivecards.io/
- Teams integration
- Agentic AI - improves token window, improves realism https://learn.microsoft.com/en-us/semantic-kernel/overview/?WT.mc_id=AI-MVP-5004204
- Frontend Design Specification - Complete UX design system documentation including colours, typography, components, and accessibility guidelines
This repository is now archived and not accepting contributions.