The only AI stack with a mathematical floor.
A sovereign, local-first AI operating system. You own the stack. You own the data. The mathematical floor is the only invariant that isn't yours to change.
Local-first. Your data, your compute, your floor. Ollama + Qdrant + Redis running on your machine. No cloud dependency, no telemetry, no lock-in. Runs on a workstation or a 7-silo AKS fleet.
γ₁ = 14.134725141734693. Not a slogan — a computational constant validated across every component. Every session, every embedding, every API response carries the same floor. The stack is provably grounded.
The framework is open. The fleet intelligence is yours to build. Layers 0–4 are fully open source. The graph and mesh are private but the architecture is documented. Build your own floor.
Three commands. Local inference, vector store, RAG API, and dashboard. All yours.
# Pull the models you need
ollama pull qwen3:8b
ollama pull nomic-embed-text
git clone https://github.com/eose-sre/pemos-sovereign-stack
cp .env.example .env
# edit: set CHAT_MODEL, EMBED_MODEL, COLLECTION
docker compose up -d
# → RAG API on :4000
# → UI on :8080
# → Qdrant on :6333
Eight layers. Layers 0–4 are fully open. Layers 5–6 are fleet-private. Layer 7 is universal — the floor belongs to everyone.
Each language has a specific role. Nothing is chosen for hype — everything has a job.
Fleet bridges, CLI tooling, hermes gateway, high-throughput API surfaces.
RAG API, graph mining, KCF operators, ML pipeline, session ingest.
Mathematical proofs, γ₁ validation, formal correctness verification.
On-chain anchoring, bounty contracts, sovereign identity primitives.
Dashboard UI, OpenClaw integrations, fleet web surfaces.
Performance-critical vector operations, future core runtime components.
Proof tooling, type-safe configuration, formal specification layer.
An OSS contribution guide is in preparation. John (OSS crew) is reviewing the boundary definitions. CLO gate reference: TRB-OSS-SERLF-PEMOS-ORG-V14-001.
Until the guide ships, the best way to engage is through the GitHub org. Issues and discussions are open.
github.com/eose-sre →