Scalable multi‑agent orchestration with persistent sessions, shared vector memory, Redis event bus, and self‑repair.
Single agents hit context and coordination limits. Complex workflows require swarms that can share memory, communicate reliably, and improve from experience.
XANDER Hive is the orchestration backbone for reliable autonomous agent teams:
- Persistent sessions — agents keep identity across tasks
- Shared vector memory — semantic recall for all agents
- Redis event bus — reliable pub/sub messaging
- RLHF‑style learning loops — agents learn from outcomes
- Self‑repair — peers can fix broken agents via review
Used across the HiveSec ecosystem but generic enough for trading bots, research swarms, etc.
flowchart LR
P[Planner Agent] --> Q[Task Queue]
Q --> E[Executor Agent]
E --> V[Vector Memory]
E --> R[Redis Event Bus]
R --> Rv[Reviewer Agent]
Rv -->|approve| D[Deliver / Deploy]
Rv -->|reject| E
V --> E
V --> Rv
D --> L[(Postgres Logs)]
subgraph "Persistence"
V
L
end
git clone https://github.com/GBOYEE/xander-hive-framework.git
cd xander-hive-framework
cp .env.example .env
docker compose up -d| Component | Role |
|---|---|
| Planner | Breaks high‑level goals into executable steps |
| Executor | Runs tools & skills, writes results |
| Reviewer | Validates outputs before delivery |
| Vector Memory | FAISS + embeddings for semantic recall |
| Event Bus | Redis pub/sub for async coordination |
| Dashboard | Streamlit UI for monitoring & control |
Inspired by multi‑agent RLHF and hive‑mind architectures. Agents operate with persistent identity, share a common memory, and learn from each other’s successes and failures. Safety is baked in via mandatory review and rollback hooks.
XANDER Hive is designed for domains where reliability and auditability matter: AIOps, security auditing, and infrastructure automation.
MIT