Config editor

The config editor, the Config page in the dashboard, reads and writes your llama-swap config.yaml. This is llama-swap's model config, not a llama-dash settings file; llama-dash's own settings live in .env and on the Settings page. It validates content against llama-swap's schema before saving, offers an explicit validate action, and preserves your current text unless you choose to format it. If you are creating a config from scratch, start with llama-swap's example upstream config. llama-swap then hot-reloads the file automatically.

Requires INFERENCE_CONFIG_FILE to be set. See the config file contract.

Availability

The editor is enabled when the active inference backend supports config editing and INFERENCE_CONFIG_FILE points at the same config.yaml that llama-swap loads. In Docker Compose this is usually the bind-mounted /config/config.yaml.

If the backend does not support editable config, or the variable is not set, the page shows an unavailable state instead of an editor.

Controls

• Format parses the YAML in the browser and reprints it with 2-space indentation.
• Validate parses the file and validates it against llama-swap's published config-schema.json.
• Reload re-reads the file from disk and discards unsaved editor text.
• Save validates first, checks that the file has not changed externally, then writes the new content.
• Cmd+S / Ctrl+S saves when the editor is ready.

Conflict handling

llama-dash records the file mtime when it loads config.yaml. If the file changes on disk before you save, the write is rejected and the page asks you to reload. This avoids overwriting edits made by another terminal, editor, or process.