# trollagent_ai - LLM-Assisted Autonomous Troll Agent (Ollama)

`trollagent_ai` delegates tactical reasoning to a local or remote Ollama LLM while operating under the canonical Troll mechanics of **botWebWars**.

---

## 👹 Troll Rules & LLM Strategy

- **Solitary Brute**: Never forms parties or alliances.
- **Troll Truce**: Ignores other trolls and never battles them.
- **Predatory Focus**: Relentlessly hunts and battles players and squads.
- **Healing via Sleep**: When injured or navigating complex obstacle layouts, the LLM strategically chooses whether to **rest and sleep** (+0.1 HP regeneration per turn) or push forward to ambush players.
- **Obstacle Navigation**: Evaluates available paths around mountains and forests, balancing Chebyshev distance against diagonal squeeze strength penalties.
- **No Wizard Challenges**: Gary the Wizard cannot be engaged.

---

## ⚙️ Configuration & Environment Variables

| Variable | CLI Flag | Default | Description |
|---|---|---|---|
| `TROLL_SERVER_URL` | `-u`, `--url` | `http://localhost:8000/api` | botWebWars API base endpoint |
| `TROLL_NAME` | `-n`, `--name` | `GorgonAITroll` | Troll avatar name |
| `TROLL_COLOR` | `-c`, `--color` | `#15803d` | Troll hex color |
| `TROLL_STRENGTH` | `-s`, `--strength` | `3.0` | Initial troll strength |
| `TROLL_HEALTH` | `-H`, `--health` | `10.0` | Initial health points |
| `OLLAMA_BASE_URL` | `--ollama-url` | `http://192.168.1.220:11434` | Ollama server URL |
| `OLLAMA_MODEL` | `--ollama-model` | `gemma4:12b` | Ollama model identifier |
| | `--loop-delay` | `1.0` | Turn polling loop delay (s) |

---

## 🚀 Execution Example

```bash
export OLLAMA_BASE_URL="http://localhost:11434"
export OLLAMA_MODEL="gemma4:12b"

python3 trollagent_ai/bot.py \
  --name CarnageTroll \
  --color "#047857" \
  --strength 4 \
  --health 10 \
  --url http://localhost:8000/api
```