platform/src/servers/api/chat/opencode/server-manager.ts

import type { Subprocess } from 'bun';
import { homedir } from 'os';
import { join } from 'path';
import { mkdirSync } from 'fs';
import { logger } from '../logger';

// OpenCode binds a session to the SERVER's working directory (POST /session has no per-session
// `directory`). So a chat that must run in a specific dir needs a server rooted there:
//   - general /chat (no specific cwd) → the fixed, pm2-managed server on OPENCODE_SERVER_URL.
//   - context-scoped chats (email account dir, project dir, …) → a pooled server spawned in that cwd.

export const OPENCODE_SERVER_URL = process.env.OPENCODE_SERVER_URL || 'http://127.0.0.1:4096';
const OPENCODE_BIN = process.env.OPENCODE_BIN || join(homedir(), '.opencode', 'bin', 'opencode');

const HEALTH_TIMEOUT_MS = 20_000;
const HEALTH_POLL_MS = 200;
const START_ATTEMPTS = 3;

type OpenCodeServer = { baseUrl: string; proc: Subprocess };

// Pooled per-cwd servers (one per working directory). The fixed server is not in here.
const servers = new Map<string, OpenCodeServer>();
const starting = new Map<string, Promise<OpenCodeServer>>();

// On startup, reap orphaned per-cwd serves left by a previous officer instance (each ~0.5GB) so they
// don't accumulate across restarts. The fixed server (OPENCODE_SERVER_URL) is preserved by port.
(function reapOrphanServers() {
  const fixedPort = (() => {
    try {
      return new URL(OPENCODE_SERVER_URL).port || '4096';
    } catch {
      return '4096';
    }
  })();
  try {
    Bun.spawnSync([
      'bash',
      '-c',
      `ss -tlnp 2>/dev/null | grep -F opencode | grep -oP '127\\.0\\.0\\.1:\\K[0-9]+' | sort -u | while read -r p; do ` +
        `[ "$p" = "${fixedPort}" ] && continue; ` +
        `pid=$(ss -tlnp 2>/dev/null | grep "127.0.0.1:$p " | grep -oP 'pid=\\K[0-9]+' | head -1); ` +
        `[ -n "$pid" ] && kill "$pid" 2>/dev/null; done`,
    ]);
  } catch {
    /* best effort */
  }
})();

async function isHealthy(baseUrl: string): Promise<boolean> {
  try {
    const res = await fetch(`${baseUrl}/api/health`, { signal: AbortSignal.timeout(3000) });
    return res.ok;
  } catch {
    return false;
  }
}

export async function isServerHealthy(): Promise<boolean> {
  return isHealthy(OPENCODE_SERVER_URL);
}

async function waitHealthy(baseUrl: string, timeoutMs: number): Promise<boolean> {
  const deadline = Date.now() + timeoutMs;
  while (Date.now() < deadline) {
    if (await isHealthy(baseUrl)) return true;
    await new Promise((r) => setTimeout(r, HEALTH_POLL_MS));
  }
  return false;
}

/** Grab an ephemeral free port by briefly binding one and releasing it. */
function getFreePort(): number {
  const probe = Bun.serve({ port: 0, hostname: '127.0.0.1', fetch: () => new Response('') });
  const port = probe.port;
  probe.stop(true);
  if (port == null) throw new Error('failed to acquire a free port');
  return port;
}

async function startServer(cwd: string, home: string): Promise<OpenCodeServer> {
  mkdirSync(cwd, { recursive: true });

  for (let attempt = 1; attempt <= START_ATTEMPTS; attempt += 1) {
    const port = getFreePort();
    const baseUrl = `http://127.0.0.1:${port}`;
    // Loopback-only + no password → open on 127.0.0.1. HOME is the user's home so `opencode` reads
    // that user's ~/.local/share/opencode auth. cwd roots the session working directory.
    const proc = Bun.spawn([OPENCODE_BIN, 'serve', '--port', String(port), '--hostname', '127.0.0.1'], {
      cwd,
      env: { ...process.env, HOME: home },
      stdout: 'ignore',
      stderr: 'ignore',
    });

    if (await waitHealthy(baseUrl, HEALTH_TIMEOUT_MS)) {
      logger.info('opencode serve started', { cwd, baseUrl });
      proc.exited.then((code) => {
        if (servers.get(cwd)?.proc === proc) servers.delete(cwd);
        logger.warn('opencode serve exited', { cwd, code });
      });
      return { baseUrl, proc };
    }

    logger.warn('opencode serve failed health check, retrying', { cwd, baseUrl, attempt });
    try {
      proc.kill();
    } catch {
      /* already gone */
    }
  }

  throw new Error(`opencode serve failed to start for cwd ${cwd}`);
}

async function ensureCwdServer(cwd: string, home: string): Promise<{ baseUrl: string }> {
  const existing = servers.get(cwd);
  if (existing && (await isHealthy(existing.baseUrl))) return { baseUrl: existing.baseUrl };
  if (existing) {
    try {
      existing.proc.kill();
    } catch {
      /* already gone */
    }
    servers.delete(cwd);
  }

  const inflight = starting.get(cwd);
  if (inflight) return { baseUrl: (await inflight).baseUrl };

  const promise = startServer(cwd, home);
  starting.set(cwd, promise);
  try {
    const server = await promise;
    servers.set(cwd, server);
    return { baseUrl: server.baseUrl };
  } finally {
    starting.delete(cwd);
  }
}

/**
 * The base URL of the OpenCode server a chat should use. With no cwd → the fixed pm2 server; with a
 * cwd → a pooled server rooted at that directory (so the session's agent operates there).
 */
export async function ensureServer(cwd?: string, home?: string): Promise<{ baseUrl: string }> {
  if (!cwd) return { baseUrl: OPENCODE_SERVER_URL };
  return ensureCwdServer(cwd, home ?? homedir());
}