SIIMPAF

Synthetic Intelligence Interactive Matrix Personal Adaptive Familiar

Synthetic Intelligence Interactive Matrix Personal Adaptive Familiar
Pre-Alpha v0.0.38 Self-Hosted AI Open Source

About SIIMPAF

SIIMPAF is a comprehensive self-hosted AI infrastructure that provides document processing, vector-based search, and intelligent content analysis capabilities using consumer hardware and open-source tools. The name bridges modern technology with the mystical concept of a familiar - a helpful companion that assists in daily tasks.

The system consists of a Python backend with FastAPI REST API server and integrates with tools like Ollama, vLLM, and Qdrant for private, self-hosted AI resources.

Key Features

Document Processing

Process and analyze documents with AI-powered extraction, summarization, and semantic search.

Vector Search (RAG)

Retrieval-augmented generation using Qdrant vector database for context-aware responses.

Avatar Animation

Animated AI avatars using SadTalker, EMAGE, and PantoMatrix for lip-sync and body motion.

Multi-LLM Support

Supports Ollama, vLLM, and LoRA fine-tuned models for different use cases.

Text-to-Speech

High-quality voice synthesis with Coqui TTS, Edge TTS, and voice cloning support.

Distributed Computing

GPU workloads distributed across DGPUNET Ray cluster for demanding tasks.

Technology Stack

Python FastAPI Ollama vLLM Qdrant PyTorch Stable Diffusion Ray Cluster CUDA 12.x Coqui TTS

AI/LLM Tools Used

Large Language Models (LLMs)

Vector Database & RAG

Image Generation

Animation Pipeline

Text-to-Speech & Speech Recognition

Distributed Computing

Practical Application: RPEPTFS

Role-playing Enhanced Pitch Training Feedback Simulator

RPEPTFS demonstrates SIIMPAF's capabilities in a real-world application: an AI-powered pitch training platform where entrepreneurs practice their investor presentations with realistic AI-generated investor NPCs.

Each investor NPC has a distinct personality, investment focus, and questioning style powered by QLoRA fine-tuned LLMs. The system generates animated avatar responses complete with lip-sync, facial expressions, and body gestures - creating an immersive training environment that prepares founders for actual investor meetings.

RPEPTFS Features Powered by SIIMPAF:

  • 5-6 simultaneous AI investor NPCs with unique personalities
  • Real-time animated avatar responses via DGPUNET cluster
  • LoRA fine-tuned models for realistic investor behavior
  • Vector-based context retention for coherent conversations
  • Text-to-speech with Coqui TTS voice synthesis
  • Full-body animation using EMAGE + PantoMatrix pipeline
🎯
5-6 NPCs
Simultaneous Investors
🎭
Animated
Full-Body Avatars
🧠
Fine-Tuned
Investor Personalities

Why DGPUNET is Essential: Running 5-6 simultaneous AI investors with real-time animation would overwhelm a single GPU. The nv5090's 32GB VRAM can handle only 2 NPCs at once. By distributing the workload across DGPUNET's 92GB of total VRAM, RPEPTFS can deliver the full multi-NPC experience that makes pitch training realistic and valuable.

Other Applications

SIIMPAF's modular architecture enables diverse applications beyond RPEPTFS:

📚

LMS Integration

AI tutoring assistants for learning management systems with context-aware responses.

🎵

Music Education

Intelligent practice companions for Practicing Musician's platform.

🎲

Therapeutic Gaming

AI-powered NPCs for therapeutic role-playing game sessions.

📄

Document Analysis

RAG-powered research assistants for large document repositories.

Related Articles

The following articles document the journey and philosophy behind SIIMPAF and related AI infrastructure:

Building DGPUNET: Democratizing AI Innovation Through Open Source Infrastructure

October 2025

A philosophical statement about accessibility in AI development - building distributed GPU infrastructure using consumer hardware when cloud providers gatekeep resources.

Part 1: How Role-Playing Games and Early Computing Shaped Four Decades of AI Development (1977-1990)

October 2025 | ~12 min read

Early experiences with pattern matching, NPC behaviors, and making computers feel responsive laid the foundation for four decades of AI development.

Part 2: IRC Bots, Beowulf Clusters, and Distributed Computing (1990-2005)

October 2025 | ~15 min read

IRC bots for natural language processing, Beowulf clusters, building ISPs and data centers - commodity over enterprise, distributed over centralized.

Part 3: Professional Applications and Breakthroughs (2005-2020)

October 2025

Therapeutic gaming, educational technology, and real-time AI systems that outperformed commercial solutions.

Part 4: Building Distributed GPU Infrastructure When Centralization Threatens Innovation (2020-2025)

October 2025

GPU scarcity, centralization concerns, and building DGPUNET to democratize AI computational resources.

Part 5: SIIMPAF - Four Decades of Technology Lessons in One System

October 2025

Bringing four decades of lessons together in SIIMPAF - a comprehensive self-hosted AI system embodying distributed computing and computational independence.

How Role-Playing Games, Distributed Computing, and AI Led to an Investment Pitch Simulator

October 2025

From IRC bots and NPC behaviors to RPEPTFS - an AI-powered pitch training platform with QLoRA-trained investor NPCs.

Related Projects