docs/devdoc/mainloop.md

### Step 1: Define Multiple Players with LLM Backend

```python
from agentreview.agent import Player
from agentreview.backends import OpenAIChat

# Describe the environment (which is shared by all players)
environment_description = "It is in a university classroom ..."

# A "Professor" player
player1 = Player(name="Professor", backend=OpenAIChat(),
                 role_desc="You are a professor in ...",
                 global_prompt=environment_description)
# A "Student" player
player2 = Player(name="Student", backend=OpenAIChat(),
                 role_desc="You are a student who is interested in ...",
                 global_prompt=environment_description)
# A "Teaching Assistant" player
player3 = Player(name="Teaching assistant", backend=OpenAIChat(),
                 role_desc="You are a teaching assistant of the module ...",
                 global_prompt=environment_description)
```

### Step 2: Create a Language Game Environment

You can also create a language model-driven environment and add it to the ChatArena:

```python
from agentreview.environments.conversation import Conversation

env = Conversation(player_names=[p.name for p in [player1, player2, player3]])
```

### Step 3: Run the Language Game using Arena

`Arena` is a utility class to help you run language games:

```python
from agentreview.arena import Arena

arena = Arena(players=[player1, player2, player3],
              environment=env, global_prompt=environment_description)
# Run the game for 10 steps
arena.run(num_steps=10)

# Alternatively, you can run your own main loop
for _ in range(10):
    arena.step()
    # Your code goes here ...
```

You can easily save your gameplay history to file:

```python
arena.save_history(path=...)
```

and save your game config to file:

```python
arena.save_config(path=...)
```