storkit/server/src/main.rs

mod agent_log;
mod agents;
mod config;
mod http;
mod io;
mod llm;
pub mod log_buffer;
mod state;
mod store;
mod workflow;
mod worktree;

use crate::agents::AgentPool;
use crate::http::build_routes;
use crate::http::context::AppContext;
use crate::http::{remove_port_file, resolve_port, write_port_file};
use crate::io::fs::find_story_kit_root;
use crate::state::SessionState;
use crate::store::JsonFileStore;
use crate::workflow::WorkflowState;
use poem::Server;
use poem::listener::TcpListener;
use std::path::PathBuf;
use std::sync::Arc;
use tokio::sync::broadcast;

#[tokio::main]
async fn main() -> Result<(), std::io::Error> {
    let app_state = Arc::new(SessionState::default());
    let cwd = std::env::current_dir().unwrap_or_else(|_| PathBuf::from("."));
    let store = Arc::new(
        JsonFileStore::from_path(PathBuf::from("store.json")).map_err(std::io::Error::other)?,
    );

    // Auto-detect a .story_kit/ project in cwd or parent directories.
    if let Some(project_root) = find_story_kit_root(&cwd) {
        io::fs::open_project(
            project_root.to_string_lossy().to_string(),
            &app_state,
            store.as_ref(),
        )
        .await
        .unwrap_or_else(|e| {
            slog!("Warning: failed to auto-open project at {project_root:?}: {e}");
            project_root.to_string_lossy().to_string()
        });

        // Validate agent config for the detected project root.
        config::ProjectConfig::load(&project_root)
            .unwrap_or_else(|e| panic!("Invalid project.toml: {e}"));
    } else {
        // No .story_kit/ found — fall back to cwd so existing behaviour is preserved.
        *app_state.project_root.lock().unwrap() = Some(cwd.clone());
    }

    let workflow = Arc::new(std::sync::Mutex::new(WorkflowState::default()));
    let port = resolve_port();

    // Filesystem watcher: broadcast channel for work/ pipeline changes.
    // Created before AgentPool so the pool can emit AgentStateChanged events.
    let (watcher_tx, _) = broadcast::channel::<io::watcher::WatcherEvent>(1024);
    let agents = Arc::new(AgentPool::new(port, watcher_tx.clone()));

    // Start the background watchdog that detects and cleans up orphaned Running agents.
    // When orphans are found, auto-assign is triggered to reassign free agents.
    let watchdog_root: Option<PathBuf> = app_state.project_root.lock().unwrap().clone();
    AgentPool::spawn_watchdog(Arc::clone(&agents), watchdog_root);
    if let Some(ref root) = *app_state.project_root.lock().unwrap() {
        let work_dir = root.join(".story_kit").join("work");
        if work_dir.is_dir() {
            io::watcher::start_watcher(work_dir, root.clone(), watcher_tx.clone());
        }
    }

    // Subscribe to watcher events so that auto-assign triggers when a work item
    // file is moved into an active pipeline stage (2_current/, 3_qa/, 4_merge/).
    {
        let watcher_auto_rx = watcher_tx.subscribe();
        let watcher_auto_agents = Arc::clone(&agents);
        let watcher_auto_root: Option<PathBuf> =
            app_state.project_root.lock().unwrap().clone();
        if let Some(root) = watcher_auto_root {
            tokio::spawn(async move {
                let mut rx = watcher_auto_rx;
                while let Ok(event) = rx.recv().await {
                    if let io::watcher::WatcherEvent::WorkItem { ref stage, .. } = event
                        && matches!(stage.as_str(), "2_current" | "3_qa" | "4_merge")
                    {
                        slog!(
                            "[auto-assign] Watcher detected work item in {stage}/; \
                             triggering auto-assign."
                        );
                        watcher_auto_agents
                            .auto_assign_available_work(&root)
                            .await;
                    }
                }
            });
        }
    }

    // Reconciliation progress channel: startup reconciliation → WebSocket clients.
    let (reconciliation_tx, _) =
        broadcast::channel::<agents::ReconciliationEvent>(64);

    // Permission channel: MCP prompt_permission → WebSocket handler.
    let (perm_tx, perm_rx) = tokio::sync::mpsc::unbounded_channel();

    // Capture project root, agents Arc, and reconciliation sender before ctx
    // is consumed by build_routes.
    let startup_root: Option<PathBuf> = app_state.project_root.lock().unwrap().clone();
    let startup_agents = Arc::clone(&agents);
    let startup_reconciliation_tx = reconciliation_tx.clone();
    // Clone for shutdown cleanup — kill orphaned PTY children before exiting.
    let agents_for_shutdown = Arc::clone(&agents);

    let ctx = AppContext {
        state: app_state,
        store,
        workflow,
        agents,
        watcher_tx,
        reconciliation_tx,
        perm_tx,
        perm_rx: Arc::new(tokio::sync::Mutex::new(perm_rx)),
    };

    let app = build_routes(ctx);

    // Background reaper: periodically remove completed/failed agent entries
    // that have exceeded the TTL.
    {
        let reaper_agents = Arc::clone(&startup_agents);
        let ttl = std::time::Duration::from_secs(agents::DEFAULT_AGENT_TTL_SECS);
        tokio::spawn(async move {
            // Check every 5 minutes.
            let interval = std::time::Duration::from_secs(300);
            loop {
                tokio::time::sleep(interval).await;
                reaper_agents.reap_expired_agents(ttl);
            }
        });
    }

    // On startup:
    // 1. Reconcile any stories whose agent work was committed while the server was
    //    offline (worktree has commits ahead of master but pipeline didn't advance).
    // 2. Auto-assign free agents to remaining unassigned work in the pipeline.
    if let Some(root) = startup_root {
        tokio::spawn(async move {
            slog!(
                "[startup] Reconciling completed worktrees from previous session."
            );
            startup_agents
                .reconcile_on_startup(&root, &startup_reconciliation_tx)
                .await;
            slog!(
                "[auto-assign] Scanning pipeline stages for unassigned work."
            );
            startup_agents.auto_assign_available_work(&root).await;
        });
    }
    let addr = format!("127.0.0.1:{port}");

    println!(
        "\x1b[95;1m  ____  _             _    ___ _   \n / ___|| |_ ___  _ __| | _|_ _| |_ \n \\___ \\| __/ _ \\| '__| |/ /| || __|\n  ___) | || (_) | |  |   < | || |_ \n |____/ \\__\\___/|_|  |_|\\_\\___|\\__|\n\x1b[0m"
    );
    println!("STORYKIT_PORT={port}");
    println!("\x1b[96;1mFrontend:\x1b[0m \x1b[94mhttp://{addr}\x1b[0m");
    println!("\x1b[92;1mOpenAPI Docs:\x1b[0m \x1b[94mhttp://{addr}/docs\x1b[0m");

    let port_file = write_port_file(&cwd, port);

    let result = Server::new(TcpListener::bind(&addr)).run(app).await;

    // Kill all active PTY child processes before exiting to prevent orphaned
    // Claude Code processes from running after the server restarts.
    agents_for_shutdown.kill_all_children();

    if let Some(ref path) = port_file {
        remove_port_file(path);
    }

    result
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    #[should_panic(expected = "Invalid project.toml: Duplicate agent name")]
    fn panics_on_duplicate_agent_names() {
        let tmp = tempfile::tempdir().unwrap();
        let sk = tmp.path().join(".story_kit");
        std::fs::create_dir_all(&sk).unwrap();
        std::fs::write(
            sk.join("project.toml"),
            r#"
[[agent]]
name = "coder"

[[agent]]
name = "coder"
"#,
        )
        .unwrap();

        config::ProjectConfig::load(tmp.path())
            .unwrap_or_else(|e| panic!("Invalid project.toml: {e}"));
    }
}