Self-Hosting ZELL
April 5, 2026 ยท View on GitHub
This guide covers a production-oriented deployment pattern for ZELL.
What ZELL Is
ZELL is a self-hosted knowledge intelligence and multi-agent simulation platform. You run the stack on your infrastructure, control your model endpoints, and keep your data in your own environment.
Deployment Modes
Local split services (developer mode)
Backend:
cd backend
uv sync --all-groups
uv run uvicorn main:app --reload --host 0.0.0.0 --port 8000
Frontend:
cd frontend
npm ci
npm run dev
Required Runtime Inputs
Set these in your deployment environment:
LLM_PROVIDERLLM_BASE_URLLLM_MODELLLM_TIMEOUTLLM_MAX_TOKENSLLM_TEMPERATURELLM_TOP_PCORS_ORIGINS
Optional tuning:
BOOTSTRAP_PROFILE_COUNTPOST_BOOTSTRAP_PROFILE_COUNTBOOTSTRAP_AGENT_LIMITSEMANTIC_SCAN_MAX_RESPONSES
Security Checklist
- Put backend behind TLS (Nginx/Caddy/Traefik)
- Restrict
CORS_ORIGINSto trusted domains - Avoid exposing internal LLM endpoints publicly
- Run with least-privilege container/user permissions
- Scan dependencies regularly
Persistence and Data
Current backend data paths include:
backend/agents.dbbackend/agents_data/
For production, mount persistent volumes for both.
Health and Smoke Tests
Check service status:
curl http://localhost:8000/health
curl http://localhost:8000/api/llm/health
Bootstrap world data:
curl -X POST http://localhost:8000/api/bootstrap \
-H "Content-Type: application/json" \
-d '{"count": 1200, "with_agents": true}'
Start a simulation:
curl -X POST http://localhost:8000/api/simulation/start \
-H "Content-Type: application/json" \
-d '{"event": "Regional climate shock", "cycles": 2, "year": 2026}'
List runs:
curl http://localhost:8000/api/dashboard/runs
Scaling Notes
- Horizontal API scale should be paired with shared/persistent storage
- For larger workloads, move SQLite to a network database
- Keep LLM latency under control with model sizing and queueing
- Rate-limit expensive generation endpoints in public environments
Observability
Recommended additions:
- Reverse-proxy access logs
- Structured backend logs
- Metrics collection (CPU, memory, request latency)
- Uptime checks for
/healthand key API routes
CI/CD
Repository workflows run checks on push and pull request for both backend and frontend quality gates.
See:
.github/workflows/backend.yml.github/workflows/frontend.yml