🛡️ Jev Tibia Autonomous Bot

An autonomous game-playing decision engine for Tibia (Open Tibia / OTServ) powered by TypeSafe AI's Jev System One model.

Unlike conversational LLMs that take 2-3 seconds to generate text, Jev returns typed probabilistic judgments in ~250ms, making it fast enough to keep pace with Tibia's 200ms global cooldowns, monster combos, and kiting mechanics.

---

🏛️ Architecture (Approach A: OTClient Lua Bridge)

┌────────────────────────────────────────────────────────┐
│               OTClient (Tibia Client)                  │
│                                                        │
│  lua/jev_bot/jev_bot.lua                               │
│  - Polls g_game and g_map every 250ms                  │
│  - Packages HP, Mana, Position, Monsters into JSON     │
│  - Sends POST http://localhost:3000/tick               │
└───────────────────────────┬────────────────────────────┘
                            │ HTTP POST (Game Snapshot)
                            ▼
┌────────────────────────────────────────────────────────┐
│           Node.js Bot Server (src/server.ts)           │
│                                                        │
│  src/engine/jevTibiaBrain.ts                           │
│  - Parallel Jev System One questions:                  │
│    1. Survival & Healing (choice)                      │
│    2. Critical Lethal Danger (noul)                    │
│    3. Target Prioritization (choice among visible mobs)│
│    4. Offensive Spell/Rune (choice)                    │
│    5. Kiting & Tactical Movement (choice)              │
│  - Translates decisions into executable actions        │
└───────────────────────────┬────────────────────────────┘
                            │ HTTP Response (Actions JSON)
                            ▼
┌────────────────────────────────────────────────────────┐
│               OTClient Execution                       │
│  - g_game.talk("exura vita")                           │
│  - g_game.attack(creature)                             │
│  - g_game.useInventoryItemWith(3155, target)           │
│  - g_game.walk(direction)                              │
└────────────────────────────────────────────────────────┘

---

🚀 Getting Started

1. Test Without OTClient (Mock Game Harness)

You can immediately watch Jev fight in an Ankrahmun Dragon Lair simulation directly in your terminal:

npm run simulate

2. Run Automated Tests

Verifies critical HP emergency healing, wounded target prioritization, and mana recovery:

npm test

3. Run the Live Server

Start the server listening for ticks from OTClient on port 3000:

npm start

---

🎮 Connecting to OTClient (Real Server)

1. Locate your OTClient directory (e.g. otclient-v8 or standard otclient).
2. Copy the module:

   cp -r lua/jev_bot <your_otclient_path>/modules/

3. Launch OTClient and log in to your OTServ.
4. Click the "Jev AI Bot" button on the top menu (or open the module) and click "Start Jev Bot".
5. The bot will begin streaming ticks to http://localhost:3000/tick and executing Jev's decisions!

---

🧠 Decisions Evaluated Every Tick

In every 250ms cycle, Jev evaluates:

• survival_action (choice): Selects cast_exura_vita, cast_exura_gran, cast_exura, use_health_potion, use_mana_potion, cast_utamo_vita, or none.
• is_critical_danger (noul): Probabilistic judgment of whether incoming burst damage is lethal.
• target_selection (choice): Analyzes all visible monsters in the battle list and targets the optimal mob (e.g. wounded dragons to quickly reduce incoming damage).
• offensive_action (choice): Decides whether to cast single-target strike spells (exori frigo), Sudden Death runes (SD), area runes (GFB / Avalanche), or auto-attack.
• tactical_movement (choice): Decides whether to kite backwards, diagonal-step to dodge monster breath/waves, or hold ground.