Ω
OmniSmart Suite Open Source 2026

Sovereign, High-Performance
Open-Source AI Platform

Six dual-licensed (AGPLv3 / Commercial) systems engineered for real-world enterprise workloads. Explore live interactive demos, benchmarks, and architectural specifications.

Sovereign RAG Live Demo

IntelAI

Persona-aware AI analytics & RAG copilot with hybrid retrieval (Dense BGE-M3 + BM25 + RRF + Cross-Encoder reranking) and Monte-Carlo forecasting.

• Metric: 4.64% Mean APE across 378 out-of-sample forecasts
• PyPI: omnismart-personas
Document AI Live Demo

DocIntel

Vision-first document intelligence with 3 execution routes (Multimodal LLM / Local Ollama / Surya OCR + Tesseract) & map-reduce for 100+ page PDFs.

• Metric: 95.0% zero-shot field accuracy on SROIE
• Stack: FastAPI + Surya OCR + Ollama Vision
Speech AI Live Demo

VoiceFlow

Low-latency streaming voice orchestration with provider-agnostic transcription (WhisperX / Deepgram / Groq), WebRTC, and meeting intelligence.

• Metric: 2.2% WER / 0.8% CER on LibriSpeech clean
• Latency: 1.157s mean WebSocket connection
MCP Runtime Live Demo

AgentKit

Governed Model Context Protocol (MCP) runtime with dynamic tool packs, typed effects (read/write/destructive), cryptographic approval tokens, & audit trails.

• Metric: 14/14 guardrails pass, 100% tool accuracy
• PyPI: agentkit-mcp
Telemetry Ingestion Live Demo

StreamPulse

Real-time data ingestion and classification pipeline with HMAC verification, 3-tier cascade classification, deduplication, & 3-sigma anomaly detection.

• Metric: 0.990 Macro-F1 across 504 events
• Connectors: n8n, Kafka, Redis, pgvector
LLMOps Evals Live Demo

RAGeval

Drop-in LLMOps observability for RAG — multi-judge consensus with disagreement alarm, 5 composite metric scorers, bootstrap CIs, and OpenTelemetry tracing.

• Metric: 0.860 accuracy on HaluEval-QA (N=200)
• PyPI: omnismart-rageval