docs: add Bannerlord VM setup guide and GABS connectivity test (#1098)

- docs/research/bannerlord-vm-setup.md: step-by-step guide for provisioning the Windows VM, installing Bannerlord + GABS mod, verifying the GABS TCP server on port 4825, and confirming connectivity from Hermes; includes reproducibility checklist - scripts/test_gabs_connectivity.py: smoke test script for GABS TCP JSON-RPC connectivity from Hermes (mirrors test_hermes4.py style); covers TCP connection, ping round-trip, get_game_state, and LAN latency target (< 100 ms) Fixes #1098 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-23 14:14:32 -04:00
parent 852fec3681
commit 3b580c4905
2 changed files with 474 additions and 0 deletions
--- a/docs/research/bannerlord-vm-setup.md
+++ b/docs/research/bannerlord-vm-setup.md
@@ -0,0 +1,230 @@
+# Bannerlord Windows VM Setup Guide
+
+**Issue:** #1098
+**Parent Epic:** #1091 (Project Bannerlord)
+**Date:** 2026-03-23
+**Status:** Reference
+
+---
+
+## Overview
+
+This document covers provisioning the Windows VM that hosts Bannerlord + GABS mod,
+verifying the GABS TCP JSON-RPC server, and confirming connectivity from Hermes.
+
+Architecture reminder:
+```
+Timmy (Qwen3 on Ollama, Hermes M3 Max)
+  → GABS TCP/JSON-RPC (port 4825)
+    → Bannerlord.GABS C# mod
+      → Game API + Harmony
+        → Bannerlord (Windows VM)
+```
+
+---
+
+## 1. Provision Windows VM
+
+### Minimum Spec
+| Resource | Minimum | Recommended |
+|----------|---------|-------------|
+| CPU | 4 cores | 8 cores |
+| RAM | 16 GB | 32 GB |
+| Disk | 100 GB SSD | 150 GB SSD |
+| OS | Windows Server 2022 / Windows 11 | Windows 11 |
+| Network | Private VLAN to Hermes | Private VLAN to Hermes |
+
+### Hetzner (preferred)
+```powershell
+# Hetzner Cloud CLI — create CX41 (4 vCPU, 16 GB RAM, 160 GB SSD)
+hcloud server create \
+  --name bannerlord-vm \
+  --type cx41 \
+  --image windows-server-2022 \
+  --location nbg1 \
+  --ssh-key your-key
+```
+
+### DigitalOcean alternative
+```
+Droplet: General Purpose 4 vCPU / 16 GB / 100 GB SSD
+Image: Windows Server 2022
+Region: Same region as Hermes
+```
+
+### Post-provision
+1. Enable RDP (port 3389) for initial setup only — close after configuration
+2. Open port 4825 TCP inbound from Hermes IP only
+3. Disable Windows Firewall for 4825 or add specific allow rule:
+   ```powershell
+   New-NetFirewallRule -DisplayName "GABS TCP" -Direction Inbound `
+     -Protocol TCP -LocalPort 4825 -Action Allow
+   ```
+
+---
+
+## 2. Install Steam + Bannerlord
+
+### Steam installation
+1. Download Steam installer from store.steampowered.com
+2. Install silently:
+   ```powershell
+   .\SteamSetup.exe /S
+   ```
+3. Log in with a dedicated Steam account (not personal)
+
+### Bannerlord installation
+```powershell
+# Install Bannerlord (App ID: 261550) via SteamCMD
+steamcmd +login <user> <pass> +app_update 261550 validate +quit
+```
+
+### Pin game version
+GABS requires a specific Bannerlord version. To pin and prevent auto-updates:
+1. Right-click Bannerlord in Steam → Properties → Updates
+2. Set "Automatic Updates" to "Only update this game when I launch it"
+3. Record the current version in `docs/research/bannerlord-vm-setup.md` after installation
+
+```powershell
+# Check installed version
+Get-Content "C:\Program Files (x86)\Steam\steamapps\appmanifest_261550.acf" |
+  Select-String "buildid"
+```
+
+---
+
+## 3. Install GABS Mod
+
+### Source
+- NexusMods: https://www.nexusmods.com/mountandblade2bannerlord/mods/10419
+- GitHub: https://github.com/BUTR/Bannerlord.GABS
+- AGENTS.md: https://github.com/BUTR/Bannerlord.GABS/blob/master/AGENTS.md
+
+### Installation via Vortex (NexusMods)
+1. Install Vortex Mod Manager
+2. Download GABS mod package from NexusMods
+3. Install via Vortex — it handles the Modules/ directory layout automatically
+4. Enable in the mod list and set load order after Harmony
+
+### Manual installation
+```powershell
+# Copy mod to Bannerlord Modules directory
+$BannerlordPath = "C:\Program Files (x86)\Steam\steamapps\common\Mount & Blade II Bannerlord"
+Copy-Item -Recurse ".\Bannerlord.GABS" "$BannerlordPath\Modules\Bannerlord.GABS"
+```
+
+### Required dependencies
+- **Harmony** (BUTR.Harmony) — must load before GABS
+- **ButterLib** — utility library
+Install via the same method as GABS.
+
+### GABS configuration
+GABS TCP server listens on `0.0.0.0:4825` by default. To confirm or override:
+```
+%APPDATA%\Mount and Blade II Bannerlord\Configs\Bannerlord.GABS\settings.json
+```
+Expected defaults:
+```json
+{
+  "ServerHost": "0.0.0.0",
+  "ServerPort": 4825,
+  "LogLevel": "Information"
+}
+```
+
+---
+
+## 4. Verify GABS TCP Server
+
+### Start Bannerlord with GABS
+Launch Bannerlord with the mod enabled. GABS starts its TCP server during game
+initialisation. Watch the game log for:
+```
+[GABS] TCP server listening on 0.0.0.0:4825
+```
+
+Log location:
+```
+%APPDATA%\Mount and Blade II Bannerlord\logs\rgl_log_*.txt
+```
+
+### Local connectivity check (on VM)
+```powershell
+# Verify port is listening
+netstat -an | findstr 4825
+
+# Quick TCP probe
+Test-NetConnection -ComputerName localhost -Port 4825
+```
+
+### Send a test JSON-RPC call
+```powershell
+$msg = '{"jsonrpc":"2.0","method":"ping","id":1}'
+$client = New-Object System.Net.Sockets.TcpClient("localhost", 4825)
+$stream = $client.GetStream()
+$writer = New-Object System.IO.StreamWriter($stream)
+$writer.AutoFlush = $true
+$writer.WriteLine($msg)
+$reader = New-Object System.IO.StreamReader($stream)
+$response = $reader.ReadLine()
+Write-Host "Response: $response"
+$client.Close()
+```
+
+Expected response shape:
+```json
+{"jsonrpc":"2.0","result":{"status":"ok"},"id":1}
+```
+
+---
+
+## 5. Test Connectivity from Hermes
+
+Use `scripts/test_gabs_connectivity.py` (checked in with this issue):
+
+```bash
+# From Hermes (M3 Max)
+python scripts/test_gabs_connectivity.py --host <VM_IP> --port 4825
+```
+
+The script tests:
+1. TCP socket connection
+2. JSON-RPC ping round-trip
+3. `get_game_state` call
+4. Response latency (target < 100 ms on LAN)
+
+---
+
+## 6. Firewall / Network Summary
+
+| Source | Destination | Port | Protocol | Purpose |
+|--------|-------------|------|----------|---------|
+| Hermes (local) | Bannerlord VM | 4825 | TCP | GABS JSON-RPC |
+| Admin workstation | Bannerlord VM | 3389 | TCP | RDP setup (disable after) |
+
+---
+
+## 7. Reproducibility Checklist
+
+After completing setup, record:
+
+- [ ] VM provider + region + instance type
+- [ ] Windows version + build number
+- [ ] Steam account used (non-personal, credentials in secrets manager)
+- [ ] Bannerlord App version (buildid from appmanifest)
+- [ ] GABS version (from NexusMods or GitHub release tag)
+- [ ] Harmony version
+- [ ] ButterLib version
+- [ ] GABS settings.json contents
+- [ ] VM IP address (update Timmy config)
+- [ ] Connectivity test output from `test_gabs_connectivity.py`
+
+---
+
+## References
+
+- GABS GitHub: https://github.com/BUTR/Bannerlord.GABS
+- GABS AGENTS.md: https://github.com/BUTR/Bannerlord.GABS/blob/master/AGENTS.md
+- NexusMods page: https://www.nexusmods.com/mountandblade2bannerlord/mods/10419
+- Parent Epic: #1091
+- Connectivity test script: `scripts/test_gabs_connectivity.py`
--- a/scripts/test_gabs_connectivity.py
+++ b/scripts/test_gabs_connectivity.py
@@ -0,0 +1,244 @@
+#!/usr/bin/env python3
+"""GABS TCP connectivity and JSON-RPC smoke test.
+
+Tests connectivity from Hermes to the Bannerlord.GABS TCP server running on the
+Windows VM. Covers:
+  1. TCP socket connection (port 4825 reachable)
+  2. JSON-RPC ping round-trip
+  3. get_game_state call (game must be running)
+  4. Latency — target < 100 ms on LAN
+
+Usage:
+    python scripts/test_gabs_connectivity.py --host 10.0.0.50
+    python scripts/test_gabs_connectivity.py --host 10.0.0.50 --port 4825 --timeout 5
+
+Refs: #1098 (Bannerlord Infra — Windows VM Setup + GABS Mod Installation)
+Epic: #1091 (Project Bannerlord)
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import socket
+import sys
+import time
+from typing import Any
+
+DEFAULT_HOST = "127.0.0.1"
+DEFAULT_PORT = 4825
+DEFAULT_TIMEOUT = 5  # seconds
+LATENCY_TARGET_MS = 100.0
+
+
+# ── Low-level TCP helpers ─────────────────────────────────────────────────────
+
+
+def _tcp_connect(host: str, port: int, timeout: float) -> socket.socket:
+    """Open a TCP connection and return the socket. Raises on failure."""
+    sock = socket.create_connection((host, port), timeout=timeout)
+    sock.settimeout(timeout)
+    return sock
+
+
+def _send_recv(sock: socket.socket, payload: dict[str, Any]) -> dict[str, Any]:
+    """Send a newline-delimited JSON-RPC request and return the parsed response."""
+    raw = json.dumps(payload) + "\n"
+    sock.sendall(raw.encode())
+
+    buf = b""
+    while b"\n" not in buf:
+        chunk = sock.recv(4096)
+        if not chunk:
+            raise ConnectionError("Connection closed before response received")
+        buf += chunk
+
+    line = buf.split(b"\n", 1)[0]
+    return json.loads(line.decode())
+
+
+def _rpc(sock: socket.socket, method: str, params: dict | None = None, req_id: int = 1) -> dict[str, Any]:
+    """Build and send a JSON-RPC 2.0 request, return the response dict."""
+    payload: dict[str, Any] = {
+        "jsonrpc": "2.0",
+        "method": method,
+        "id": req_id,
+    }
+    if params:
+        payload["params"] = params
+    return _send_recv(sock, payload)
+
+
+# ── Test cases ────────────────────────────────────────────────────────────────
+
+
+def test_tcp_connection(host: str, port: int, timeout: float) -> tuple[bool, socket.socket | None]:
+    """PASS: TCP connection to host:port succeeds."""
+    print(f"\n[1/4] TCP connection → {host}:{port}")
+    try:
+        t0 = time.monotonic()
+        sock = _tcp_connect(host, port, timeout)
+        elapsed_ms = (time.monotonic() - t0) * 1000
+        print(f"  ✓ Connected ({elapsed_ms:.1f} ms)")
+        return True, sock
+    except OSError as exc:
+        print(f"  ✗ Connection failed: {exc}")
+        print(f"  Checklist:")
+        print(f"    - Is Bannerlord running with GABS mod enabled?")
+        print(f"    - Is port {port} open in Windows Firewall?")
+        print(f"    - Is the VM IP correct? (got: {host})")
+        return False, None
+
+
+def test_ping(sock: socket.socket) -> bool:
+    """PASS: JSON-RPC ping returns a 2.0 response."""
+    print(f"\n[2/4] JSON-RPC ping")
+    try:
+        t0 = time.monotonic()
+        resp = _rpc(sock, "ping", req_id=1)
+        elapsed_ms = (time.monotonic() - t0) * 1000
+        if resp.get("jsonrpc") == "2.0" and "error" not in resp:
+            print(f"  ✓ Ping OK ({elapsed_ms:.1f} ms): {json.dumps(resp)}")
+            return True
+        print(f"  ✗ Unexpected response ({elapsed_ms:.1f} ms): {json.dumps(resp)}")
+        return False
+    except Exception as exc:
+        print(f"  ✗ Ping failed: {exc}")
+        return False
+
+
+def test_game_state(sock: socket.socket) -> bool:
+    """PASS: get_game_state returns a result (game must be in a campaign)."""
+    print(f"\n[3/4] get_game_state call")
+    try:
+        t0 = time.monotonic()
+        resp = _rpc(sock, "get_game_state", req_id=2)
+        elapsed_ms = (time.monotonic() - t0) * 1000
+        if "error" in resp:
+            code = resp["error"].get("code", "?")
+            msg = resp["error"].get("message", "")
+            if code == -32601:
+                # Method not found — GABS version may not expose this method
+                print(f"  ~ Method not available ({elapsed_ms:.1f} ms): {msg}")
+                print(f"    This is acceptable if game is not yet in a campaign.")
+                return True
+            print(f"  ✗ RPC error ({elapsed_ms:.1f} ms) [{code}]: {msg}")
+            return False
+        result = resp.get("result", {})
+        print(f"  ✓ Game state received ({elapsed_ms:.1f} ms):")
+        for k, v in result.items():
+            print(f"    {k}: {v}")
+        return True
+    except Exception as exc:
+        print(f"  ✗ get_game_state failed: {exc}")
+        return False
+
+
+def test_latency(host: str, port: int, timeout: float, iterations: int = 5) -> bool:
+    """PASS: Average round-trip latency is under LATENCY_TARGET_MS."""
+    print(f"\n[4/4] Latency test ({iterations} pings, target < {LATENCY_TARGET_MS:.0f} ms)")
+    try:
+        times: list[float] = []
+        for i in range(iterations):
+            sock = _tcp_connect(host, port, timeout)
+            try:
+                t0 = time.monotonic()
+                _rpc(sock, "ping", req_id=i + 10)
+                times.append((time.monotonic() - t0) * 1000)
+            finally:
+                sock.close()
+
+        avg_ms = sum(times) / len(times)
+        min_ms = min(times)
+        max_ms = max(times)
+        print(f"  avg={avg_ms:.1f} ms  min={min_ms:.1f} ms  max={max_ms:.1f} ms")
+
+        if avg_ms <= LATENCY_TARGET_MS:
+            print(f"  ✓ Latency within target ({avg_ms:.1f} ms ≤ {LATENCY_TARGET_MS:.0f} ms)")
+            return True
+        print(
+            f"  ✗ Latency too high ({avg_ms:.1f} ms > {LATENCY_TARGET_MS:.0f} ms)\n"
+            f"    Check network path between Hermes and the VM."
+        )
+        return False
+    except Exception as exc:
+        print(f"  ✗ Latency test failed: {exc}")
+        return False
+
+
+# ── Main ──────────────────────────────────────────────────────────────────────
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="GABS TCP connectivity smoke test")
+    parser.add_argument(
+        "--host",
+        default=DEFAULT_HOST,
+        help=f"Bannerlord VM IP or hostname (default: {DEFAULT_HOST})",
+    )
+    parser.add_argument(
+        "--port",
+        type=int,
+        default=DEFAULT_PORT,
+        help=f"GABS TCP port (default: {DEFAULT_PORT})",
+    )
+    parser.add_argument(
+        "--timeout",
+        type=float,
+        default=DEFAULT_TIMEOUT,
+        help=f"Socket timeout in seconds (default: {DEFAULT_TIMEOUT})",
+    )
+    args = parser.parse_args()
+
+    print("=" * 60)
+    print(f"GABS Connectivity Test Suite")
+    print(f"Target: {args.host}:{args.port}")
+    print(f"Timeout: {args.timeout}s")
+    print("=" * 60)
+
+    results: dict[str, bool] = {}
+
+    # Test 1: TCP connection (gate — skip remaining if unreachable)
+    ok, sock = test_tcp_connection(args.host, args.port, args.timeout)
+    results["tcp_connection"] = ok
+    if not ok:
+        _print_summary(results)
+        return 1
+
+    # Tests 2–3 reuse the same socket
+    try:
+        results["ping"] = test_ping(sock)
+        results["game_state"] = test_game_state(sock)
+    finally:
+        sock.close()
+
+    # Test 4: latency uses fresh connections
+    results["latency"] = test_latency(args.host, args.port, args.timeout)
+
+    return _print_summary(results)
+
+
+def _print_summary(results: dict[str, bool]) -> int:
+    passed = sum(results.values())
+    total = len(results)
+    print("\n" + "=" * 60)
+    print(f"Results: {passed}/{total} passed")
+    print("=" * 60)
+    for name, ok in results.items():
+        icon = "✓" if ok else "✗"
+        print(f"  {icon} {name}")
+
+    if passed == total:
+        print("\n✓ GABS connectivity verified. Timmy can reach the game.")
+        print("  Next step: run benchmark level 0 (JSON compliance check).")
+    elif not results.get("tcp_connection"):
+        print("\n✗ TCP connection failed. VM/firewall setup incomplete.")
+        print("  See docs/research/bannerlord-vm-setup.md for checklist.")
+    else:
+        print("\n~ Partial pass — review failures above.")
+
+    return 0 if passed == total else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())