server-splash

CLI wizard that probes a Linux host, sends results to a local LLM agent for analysis, and generates an HTML dashboard page linking to discovered services. Also generates optional dashboard modules (e.g., Ollama dashboard).

Requirements

• Rust toolchain (edition 2021)
• A running OpenAI-compatible API endpoint (e.g., Ollama on http://127.0.0.1:11434)
• System tools: systemctl, docker, nvidia-smi, free, hostname

Build

cargo build --release

Binary: target/release/splash

Usage

Run the wizard:

./target/release/splash

Wizard steps

1. Hostname — detected from system, can be overridden
2. Agent endpoint — auto-discovers OpenAI-compatible endpoints; user selects model
3. Host probing — runs systemctl, docker ps, free, nvidia-smi, etc.
4. Agent analysis — raw probe output sent to LLM; returns structured service list
5. Probe table — editable table where user can:
   • Toggle service inclusion (s key, Enter on INC column)
   • Edit endpoint URLs (auto re-probed on change)
   • Edit descriptions
   • Pick icons from a 30-emoji palette
   • Edit service names
   • Navigate with arrow keys / jk / tab
6. GUI addons — pick from services.toml entries (Portainer, Grafana, etc.)
7. Dashboard modules — pick from available modules in src/modules/
8. Deployment — choose output directory or copy to /var/www/html

Caching

Agent responses are cached in agent-response.json. On re-run, you are prompted to reuse the cache or re-analyze. Full session state persists in splash-state.json — loading a previous session skips directly to the probe table.

Output

• splash-server.html — main dashboard page
• splash-state.json — session database (services, endpoints, icons, module selections)
• Per-module output directories (e.g., ollama-ui/index.html)

Adding a dashboard module

Create a directory under src/modules/ containing:

my-module/
  module.json      — { name, description, default_port, icon, url_prefix }
  mod.rs           — module_info() returning DashboardModule with generate fn pointer
  generate.rs      — HTML generation logic
  template.html    — HTML template with <!-- PLACEHOLDER --> markers
  style.css        — stylesheet copied to output

Register the module in src/modules/mod.rs → all_modules().

Configuration

Persisted to ~/.config/server-splash/splash-config.toml:

agent_url = "http://127.0.0.1:11434"
hostname = "myserver"
output_dir = "./server-splash"
gui_pairs_file = "/path/to/services.toml"
glances_api_base = "http://localhost:61208"

GUI addons are defined in services.toml (TOML format, one entry per service with src_port and icon).