deploy

Deploy and manage the ScienceDiscovery stack via host processes or Docker Compose.

64|15|Updated Aug 14, 2026
One-click install
npx skills add https://github.com/openJiuwen-ai/sciencediscovery --skill deploy-openjiuwen-ai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deploy
Source: https://github.com/openJiuwen-ai/sciencediscovery/tree/main/.agents/skills/deploy
Command: npx skills add https://github.com/openJiuwen-ai/sciencediscovery --skill deploy-openjiuwen-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Deploying the ScienceDiscovery research platform involves choosing between host-process and Docker Compose modes, verifying sandbox prerequisites like bubblewrap and user namespaces, and handling ports, tokens, and process management. This Skill guides the deployment step by step without silently reconfiguring the host machine. ## Core Features & Use Cases - Guided deployment forms: Walks through Docker Compose or host-process deployment using scripts/start-stack.sh, with read-only environment detection before any changes. - Safe operations policy: Never runs sudo or installs global software on its own; reports gaps with suggested user-run commands and requires confirmation before proceeding. - Lifecycle management: Provides start, stop, restart, log, and health-check commands for both modes, including correct process-group shutdown for backgrounded host stacks and SSH port forwarding for remote hosts. - Use Case: A researcher asks to run ScienceDiscovery on a remote Linux workstation. The Skill checks Node, pnpm, uv, and bubblewrap versions, deploys via start-stack.sh, verifies /health, and reports the UI URL plus SSH tunneling instructions. ## Quick Start Ask the assistant to deploy ScienceDiscovery locally and it will first confirm whether you prefer Docker Compose or host processes, then run read-only environment checks before starting the stack.

Frequently Asked Questions about deploy

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I deploy ScienceDiscovery locally?▼

Run ./scripts/start-stack.sh --mode local from the repository root after copying .env.example to .env. This installs dependencies, builds, and starts the API on port 4310 with the runner on 127.0.0.1:4311.

Should I use Docker Compose or host processes for ScienceDiscovery?▼

Docker Compose needs only Docker Engine 24+ and Compose v2, publishing on 127.0.0.1:4310 with state in ./data. Host processes require Node 22.19+, pnpm, Python 3, uv, and bubblewrap, and run in the foreground.

Does ScienceDiscovery work on macOS or Windows with Docker Desktop?▼

No. Both deployment modes require Linux x86_64 or aarch64 with unprivileged user namespaces because the bubblewrap sandbox is mandatory. Docker Desktop on macOS and Windows is unsupported.

Why does Ctrl-C not stop my ScienceDiscovery stack?▼

Ctrl-C only reaches a foreground stack. If started with nohup, tmux, or a supervisor, you must signal the entire process group identified from the API port, not the start-stack.sh PID alone.

Why do run_python and run_shell fail even though the API is up?▼

This usually means the host restricts unprivileged user namespaces, so bubblewrap cannot create sandboxes. The documented fix is setting kernel.apparmor_restrict_unprivileged_userns to 0, which requires root.

How do I access ScienceDiscovery running on a remote server?▼

Use SSH port forwarding from your laptop, for example ssh -N -L 4310:127.0.0.1:4310 user@host, then open http://127.0.0.1:4310 locally. Only forward the API port; the runner stays loopback-only.