Prototyper Docs GitHub ↗

Backend Architecture

All Rust logic lives in src-tauri/src/lib.rs. Commands are registered in generate_handler![]; plugin permissions (e.g., shell:default, fs:default) are declared in capabilities/default.json. Missing either causes silent failure at runtime.

Command groups (44 total)

Process (10)

Command Purpose
bun_dev Start a bun dev process for the active project
bun_build Production build via bun build
bun_install Async bun install
bun_install_sync Synchronous bun install (blocks until complete)
run_shell_command Generic async shell command
run_shell_command_sync Synchronous shell command
run_shell_command_capture Shell command with stdout/stderr capture
kill_process Kill a specific process by PID
kill_all_processes Kill all spawned processes
kill_port Kill whatever is listening on a given port

File System (9)

Command Purpose
read_dir List directory entries
read_file Read a file by path
write_file Write a file by path
create_dir Create a directory
delete_file Delete a file
delete_dir Delete a directory
rename_file Rename or move
create_symlink Create a symlink
reveal_in_explorer Open the OS file browser at a path

HTTP (1)

Command Purpose
http_request Generic HTTP request (used by cloud AI providers)

AI (9)

Command Purpose
generate_completion One-shot completion (non-streaming)
generate_completion_stream Streaming completion via Tauri Channel
stop_generation_stream Cancel an in-flight stream
resolve_tool_permission Resolve a pending ToolPermission event
resolve_ask_user Resolve a pending AskUser event
resolve_ask_user_form Resolve a pending AskUserForm event
list_ollama_models List models available on the local Ollama server
save_model_presets Persist model presets
load_model_presets Load saved model presets

Bonsai (11)

Command Purpose
bonsai_start_server Start the local Bonsai image-generation server
bonsai_stop_server Stop the Bonsai server
bonsai_server_status Get current server state
bonsai_generate_image Generate an image (streaming via Channel)
bonsai_cancel_generation Cancel an in-flight generation
bonsai_list_assets List saved assets
bonsai_delete_asset Delete an asset
bonsai_get_server_config Load server config
bonsai_save_server_config Persist server config
bonsai_schedule_stop Schedule a delayed server stop
bonsai_cancel_stop Cancel a scheduled stop

Export (2)

Command Purpose
export_project Export the whole project to a zip
export_component Export a single component

Workflows (3)

Command Purpose
save_workflow Persist a workflow graph
load_workflow Load a workflow graph
list_workflows List available workflows for the project

Layout

src-tauri/src/
  lib.rs                 # App setup, plugins, generate_handler![]
  main.rs                # Thin passthrough
  commands/
    process.rs           # Bun/shell spawning
    fs.rs                # File system
    http.rs              # HTTP client
    ai.rs                # Streaming completion, tool permissions
    ai_providers.rs      # OpenAI/Claude providers
    ai_ollama.rs         # Ollama-specific logic
    export.rs            # Project/component export
    workflows.rs         # Workflow persistence
    mod.rs
  agent/                 # AI agent module (loop, executor, tools)
  sandbox/               # Linux sandbox (landlock + seccomp + bwrap)

Adding a new command

Three steps:

  1. Define the function in the appropriate commands/*.rs file with #[tauri::command].
  2. Register it in generate_handler![] in lib.rs.
  3. If it needs a plugin (shell, fs, etc.), declare the permission in capabilities/default.json.
// src-tauri/src/commands/fs.rs
#[tauri::command]
pub async fn read_file(path: String) -> Result<String, String> {
    std::fs::read_to_string(&path).map_err(|e| e.to_string())
}

// src-tauri/src/lib.rs
#[tauri::generate_handler]
pub fn handler() {
    tauri::generate_handler![
        // ...
        commands::fs::read_file,
    ];
}

What next