slimRAG Sovereign AI Knowledge Engine for the Zero-Trust Enterprise

Deploy localized, ultra-fast RAG in minutes. Keep your IP completely private with zero-copy local memory intelligence and zero-knowledge encrypted transit.

Pure Rust, end-to-end encrypted, sub-500ms real-time knowledge retrieval — no cloud dependency, no data exfiltration risk.

slimRAG — Three Pillars of Sovereign AI

Three Pillars of Sovereign AI

Purpose-built for enterprises that demand absolute data control and real-time intelligence

🛡️

Absolute Data Sovereignty

End-to-end ChaCha20-Poly1305 encryption ensures data is encrypted at the edge and decrypted only at the RAG server. The relay hub handles ciphertext only — zero knowledge by construction.

  • Zero-knowledge encrypted transit
  • Financial & healthcare grade compliance
  • Data never leaves your premises

Sub-500ms Real-Time Intelligence

Millisecond-level document awareness via kernel-level file change detection (fanotify/inotify). Hybrid search across keyword, AST graph, and vector indexes fused via RRF for maximum coverage.

  • Instant awareness of local file changes
  • Hybrid search: keyword + graph + vector
  • No lengthy cloud index rebuilds
📦

Zero-Infrastructure Bloat

Full-stack Pure Rust with zero heavy dependencies. Total binary footprint under 10MB. No cloud vector database, no heavyweight C++ runtime — deploy on edge workstations or lightweight devices.

  • <10MB binary, Pure Rust
  • No cloud vector DB required
  • Zero operational overhead

Zero-Knowledge Data Flow Architecture

Three components working in concert: edge sync, encrypted relay, sovereign RAG

📡

slimSync

Edge Agent

Watches directories, detects changes via fanotify, slices files into semantic chunks, encrypts end-to-end

🔐

slimHub

Zero-Knowledge Relay

Stateless encrypted ring buffer — stores only ciphertext and blind-IDs. Cannot decrypt any payload

🧠

slimRagSvr

Sovereign RAG Server

Decrypts, indexes, and serves hybrid search via OpenAI-compatible API. Pluggable embedding backends

All inter-component communication uses Zenoh Pub/Sub with zero-knowledge guarantees

slimRAG — Zero-Knowledge Data Flow Architecture

Figure 1: End-to-End Zero-Knowledge Data Flow and Memory-Mapped Storage Architecture

🌍

Open Core

MIT License

Core engine fully open source. Audit the code, contribute features, build your own integrations.

GitHub Repository →
🏢

Enterprise

Commercial Support

IronCurtain hardware adaptation, custom connectors, dedicated security audit, 7x24 SLA.

Request Enterprise Demo →

Proven Performance at Scale

Benchmarked on real-world codebases (tokio + axum) for cross-file dependency tracing

Retrieval Quality

Recall@10 17.5%
MRR 0.292
Cross-file coverage 97.5%

Runtime Performance

Avg ingestion per branch 48ms
Avg retrieval latency 9ms
Branch switch latency 71ms

Zero hallucination rate on indexed documents · Zero cross-branch storage growth

Deploy Your Sovereign AI Knowledge Engine

Start with the open-source core, scale with enterprise support. Your data, your infrastructure, your AI.