SkyrimNet

Overview

SkyrimNet is a cutting-edge AI integration platform for games, beginning with Skyrim. Unlike other AI projects that take an external server approach, SkyrimNet uses an in-process design as a DLL without requiring WSL or external servers. This architectural difference offers faster response times, lower system load, and a streamlined setup process.

The most advanced AI platform for gaming - transforming every NPC into a living, breathing character with their own memories, goals, and personalities.

Revolutionary Architecture 🏗️

In-Process AI Stack ⚡

• In-Process Design: Everything runs within the game's DLL - no external server or WSL required

• High Performance: Fast response times and efficient system resource usage

• Direct Memory Access: Reads game state directly from memory instead of syncing to external systems

• Simplified Setup: Straightforward installation without server configuration

• Real-time Responsiveness: Improved freshness of data and responsive AI behavior

🚀 Installation & Setup

📥 Choose Your Build

• CUDA Build: Maximum performance for NVIDIA users with in-process Whisper

• Universal Build: Broad compatibility for all hardware configurations

⚙️ Quick Start

Download from GitHub Releases

Install using your preferred mod manager

Enable SkyrimNet.esp in your load order

Launch via SKSE and visit localhost:8080

Complete the guided setup wizard with your API keys

Experience the future of gaming AI!

First launch automatically generates default configurations for your system.

System Requirements 📋

🔧 Essential Dependencies

Skyrim Script Extender (SKSE)

Address Library for SKSE Plugins

PowerOfThree's Papyrus Extender

Latest Microsoft Visual C++ Redistributable

CUDA Build Requirements (For Maximum Performance) 🚀

NVIDIA GPU with CUDA support

CUDA Toolkit 12.x - IMPORTANT: If using the CUDA build, SkyrimNet will not load without this!

Optional Dependencies 📋

UIExtensions - Required for text input and Input Wheel 🎮 Version-Specific Requirements Skyrim SE (without ESL support):

Backported Extended ESL Support (BEES) Skyrim VR:

Skyrim VR ESL Support - Use instead of BEES

Important:⚠️ CUDA Toolkit is required for the CUDA build to load properly

External API Requirements🌐

LLM Provider: OpenRouter API key (or compatible OpenAI API)

Cloud Processing: VastAI account (optional, for cloud GPU access and automatic XTTS provisioning)

Key Features 🎪

AI Capabilities 🧠

Dynamic NPC Interactions: NPCs can react to player actions, world events, and conversations in real-time

Contextual Awareness: The system maintains knowledge of recent events and uses this to inform AI responses

Smart NPC Selection: Uses targeted fast LLM prompts to determine which NPC should react to events

Streaming Responses: Supports streaming LLM responses for much faster responses and more natural conversation flow

Multi-Modal Processing: Use the best LLM for the job. Different usecases are split out to different LLM's, allowing you to use a wider range of LLM's that were previously unsuitable

Semantic Understanding: Advanced embedding models for natural conversation flow

Living Character System 🎭

3,000+ Unique Personalities: Every vanilla NPC plus popular mod characters with detailed backstories

Dynamic Relationships: Characters remember interactions, form opinions, and develop connections

Goal-Oriented Behavior: NPCs pursue personal objectives, react to successes/failures

Emergent Conversations: Characters naturally discuss events, gossip, and share knowledge

Advanced Memory Architecture 🧠

Per-Character Memories: Memories are created from a first-person, per-character perspective. Every character remembers events differently based upon their personality and perspective.

Vector-Based Recall: Semantic similarity matching for contextually relevant memories

Importance Weighting: Critical events are remembered longer and influence behavior more

Temporal Decay: Natural forgetting patterns that mirror human memory

Memory Consolidation: Background processing creates long-term behavioral patterns

Intelligent Awareness System 👁️

Realistic Perception: NPCs only know what they can see, hear, or reasonably infer. People from "downstairs" will not hear or react to your conversations.

Spatial Intelligence: Distance, obstacles, and environmental factors affect interactions

Combat Awareness: Dynamic reactions to threats, allies, and changing battle conditions

Social Context: Understanding of relationships, hierarchies, and appropriate responses

Privacy Respect: Private conversations stay private unless realistically overheard

Multi-Modal Communication 🎙️

Voice Recognition: Natural speech input with streaming transcription

Text Interface: Typical text input via UIExtensions

Facial Animation: Synchronized expressions with speech

Multiple TTS Engines: XTTS, Zonos, and Piper support for diverse voice options

Customization & Flexibility ⚙️

Customizable Prompt Templates: Uses Inja templating system for highly configurable AI behavior

Situation-Specific Models: Support for different LLM configurations based on context and needs

Profile-Based Overrides: Ability to override settings on a per-profile basis (Not exposed via the UI in the initial Beta release, coming soon)

Extensive Configuration: Fine-tune every aspect of the AI system, from dialogue detection distance to facial animation intensity

Cloud-Native Features 🌐

One-Click VastAI: Automated cloud GPU provisioning with preconfigured environments. One click setup for XTTS

Smart Instance Management: Manage and monitor your instances, with automatic TTS endpoint configuration

Cost Optimization: Dynamic resource allocation. Automatically identifies and provisions the cheapest pod based upon your GPU requirements.

Zero-Config Setup: One click "Smart Create" button to provision a preconfigured pod.

Professional-Grade Web Interface 🌐

Live Operations Dashboard 📊

Real-Time System Monitoring: Server status, uptime, version info, and GameMaster state

Live Game Data: View nearby NPCs, recent events, active short-lived events in real-time

Performance Analytics: ThreadPool statistics, task duration analysis, error tracking

API Request Monitoring: View recent LLM requests, response times, and token usage

Pinned Characters: Quick access to frequently used characters with teleportation controls

Character Management Studio 🎨

Dual Bio System: Switch between static character bios and dynamic event-driven bios

Real-Time Actor Data: Live health, stats, location, factions, and package information

AI-Powered Generation: Create character profiles from nearby actors using LLM assistance

Character Creation: Scan nearby actors and generate comprehensive bios automatically

Bio Update System: Request AI updates to character personalities with diff preview

Split View Editor: Edit bios while monitoring live actor data simultaneously

Backup Management: Automatic backup system with restore capabilities

Memory System Interface 🧠

Vector Search Testing: Test memory recall with semantic similarity matching

Memory Analytics: Statistics on memory types, importance scores, and actor distribution

Memory Generation: Automatically generate memories from recent events and conversations

Advanced Filtering: Search by actor, type, importance, content, and creation time

Memory Management: Create, edit, delete, and organize character memories

Configuration Management ⚙️

Live Config Editing: Real-time configuration changes with immediate validation

Hotkey Configuration: Visual hotkey capture with Windows VK code mapping

Variant Support: Separate configs for CUDA vs non-CUDA builds

Config Search: Find specific settings across all configuration files

Schema Validation: Built-in validation prevents configuration errors

VastAI Cloud Integration ☁️

Instance Management: Create, monitor, and manage cloud GPU instances

Smart Provisioning: One-click setup with automatic cost optimization

Live Instance Monitoring: Real-time status, logs, and resource usage

Automatic TTS Setup: Seamless XTTS endpoint configuration for running instances

Cost Tracking: Monitor usage and automatically find cheapest available pods

Built for Innovation 🔧

Modder-Friendly Architecture 🎯

Extensible API: Powerful Papyrus API for creating custom behaviors and actions. Creating new mods generally follows normal Skyrim modding paradigms.

Hot-Reloading: Modify prompts, configs, and see changes instantly

Template Engine: Powerful Inja-based prompt system with advanced features

Event Hooks: React to any game event with custom AI behaviors

Current Limitations ⚠️

Available Actions: The default list of actions that is exposed is rather limited at the moment. This will be expanded substantively in the future.

VR Keybinds: At present, VR users cannot bind their controller buttons directly to hotkeys. VR users can use this as a workaround for the time being.

Future Plans 🚀

Expanded Actions: More actions for NPC's to use. Dynamic Quests: Start and progress (some) quests automatically. Construct meaningful story arcs via automatic quest assignment. Handle things like "Meet me at the tavern later" via this system.

Dialogue Tree Analysis and Integration: Analyze and Integrate with the existing normal "dialogue trees" to expose selection through natural dialogue, and improve profile generation . Improve VR Support: Fix the VR Keybind thing, and make VR a first class citizen. It already works quite well in VR aside from the hotkeys, for which there is a workaround.

Image to Text: Add a sophisticated ITT pipeline that introduces meaningful context for the LLM.

Much, much more