Open-source infrastructure for production-ready, event-driven autonomous AI systems.
We build the foundational runtime, memory, and orchestration layers required to run AI agents as reliable, distributed systems — not demos or chatbots.
Our focus is on:
- Event-driven execution
- Multi-agent coordination
- Persistent memory
- Horizontal scalability
- Production reliability
All components are modular, framework-agnostic, and designed to operate independently or as a unified stack.
Autonomous AI agent framework for reasoning, tool use, routing, and workflow orchestration.
- Model-agnostic (via LiteLLM)
- Multi-agent patterns (routing, parallel, sequential)
- MCP + local tool support
- Streaming and background execution
- Production-proven in real workloads
→ https://github.com/omnirexflora-labs/omnicoreagent
Event-driven runtime engine that turns agents into scalable, distributed systems.
- Framework-agnostic runtime
- Redis Streams event bus
- Consumer groups and horizontal scaling
- Retries, DLQs, and observability built-in
→ https://github.com/omnirexflora-labs/OmniDaemon
Persistent, framework-agnostic memory layer for AI agents.
- Cross-session memory persistence
- Hybrid retrieval (semantic + keyword + metadata)
- Multi-tenant isolation
- Redis, PostgreSQL, MongoDB, SQLite backends
- Vector DB integrations (Qdrant, ChromaDB, MongoDB Atlas)
→ https://github.com/omnirexflora-labs/omnimemory
uv add omnicoreagent
uv add omnidaemon
uv add omnimemory- Event-driven by default
- Framework-agnostic
- Production-first
- Modular and composable
- Open-source, MIT licensed
See ARCHITECTURE.md for a high-level system overview.
