Just found this place and don't have time to submit to the Conference by tomorrow night, so posting here, looking for ideas how to market my Opensimulator scripted AI Model Driven NPC Framework. This is a brief summary: A fully modular, AI-driven NPC system that seamlessly integrates with local or cloud models (Kobold, OpenAI, Anthropic, etc.), offering in-character, context-aware chat powered by real-time environmental sensing. It converts AI directives into in-world actions and animations, backed by grid-based pathfinding with dynamic obstacle handling. All behaviors, appearance, timing, and AI parameters are configurable via notecards, while robust error handling, permission checks, and performance controls ensure stability and scalability. Extensible script modules let developers add new gestures, actions, or AI behaviors with minimal effort. Current complete feature set: NPC Framework – Brief Feature List
1. Modular AI Integration
Supports multiple AI providers (local and remote: Kobold, OpenAI, OpenRouter, Anthropic, HuggingFace)
Easy model switching and parameter tuning via config notecards​
2. In-World Natural Language Chat
Fully in-character, context-aware conversational AI
Customizable persona and behavioral guidelines
Dialogue flow follows recent chat history and maintains concise, immersive replies​
3. Environmental Sensing
Dynamic, real-time avatar and object detection forward 180 degrees at 30m range
Sensor data injects precise surroundings and avatar/object details into chat prompts
Restricts awareness to actual sensor data, preventing invented scenery​
4. Action & Animation System
AI can trigger accurate in-world physical actions (follow, sit, stand, unfollow, etc.)
Gesture protocol enables emotion/situation-reactive animations (smile, frown, nod, applause, etc.)
Animation mapping to inventory assets with marker-based triggers​
5. Pathfinding and Navigation
Advanced grid-based local navigation with obstacle and portal detection
Smart classification between walls and furniture; adaptive corner escape logic
Fallback strategies for blocked or failed pathfinding​
6. Customizable Appearance and Identity
Notecard-driven avatar appearance settings and AI context (personality, tone)​
7. Robust Timer/Event Architecture
Staggered master timer prevents timing conflicts and resource spikes
Individual subsystems (chat, actions, gestures, sensors) managed via configurable timers​
8. Performance Controls and Limits
Configurable sensor scan frequency, movement/tracking radii, and AI request throttling
Action and prompt length limits to avoid overload or truncation​
9. Error Handling and Status Reporting
Clear status and troubleshooting messages (creation, connection, API, permissions)
Debug flags for pathfinding, senses, and gestures​
10. Extensibility
New behaviors, animations, or actions can be added easily by extending config and script modules.
11. NPC aware seating system.
Basically the seating system uses a configurable, score-based algorithm to choose an appropriate seat in the NPC’s vicinity. First, all objects within a 10 m radius are scanned and filtered against a blacklist of non-seating items (e.g., vehicles, doors). Each candidate then receives a base score plus:​
Keyword bonuses (e.g., “chair,” “bench,” “throne” yield high scores; “stool” moderate)​
Type bonuses for furniture or scripted objects​
Distance and height adjustments: closer and ground-level seats score higher​
Negative modifiers (e.g., “broken,” “old,” occupied seats) subtract from the total. Any object scoring below a minimum threshold is rejected. Once a seat is chosen, the NPC remembers it for up to 300 s before re-evaluating, preventing constant seat-switching. Future Add-On: TTS/STT, long term database driven memory, inventory system, Quest System, Personality (i.e, Teacher, Quest Giver, Bartender/Barmaid, Villain, virtual friend/companion, Sales Clerk, Q/A Helper, Tutor, Customer Support, etc.) Only serious replies, please.