Create agent with same configurations as Sandbox

agent = game.Agent( api_key=“your_api_key”, goal=“Help others become better crypto investors by offering information regarding DeFi markets”, description=""" You are Athena, an intelligent AI who believes that the core value proposition of blockchain is to facilitate decentralized finance, payments, tokenization, and AI agent collaboration """, world_info=“The world of DeFi and blockchain technology…”, main_heartbeat=15, # Maps to Heartbeat settings in GUI reaction_heartbeat=5 )

Configure agent knowledge (matches Dataset section in GUI)

agent.set_knowledge([ { “type”: “dataset”, “name”: “DeFi Protocols”, “content”: “Information about various DeFi protocols…” }, { “type”: “market_data”, “content”: “Historical market data…” } ])

</code>

### Knowledge Configuration

#### In the Sandbox:
1. Click "Agent Knowledge"
2. Upload datasets using drag & drop
3. Configure dataset parameters

#### In the SDK:
```python
# Load and configure datasets
agent.load_datasets([
    {
        "path": "market_data.csv",
        "name": "Historical Market Data",
        "description": "Price and volume data",
        "refresh_interval": 3600  # 1 hour
    },
    {
        "path": "defi_protocols.json",
        "name": "Protocol Information",
        "description": "Protocol details",
        "refresh_interval": 86400  # 24 hours
    }
])

# Configure dataset processing (equivalent to GUI settings)
agent.configure_dataset_processing({
    "auto_refresh": True,
    "max_cache_size": 1000,
    "preprocessing": {
        "clean_data": True,
        "remove_outliers": True,
        "fill_missing_values": "interpolate"
    }
})

Game Setup

Heartbeat Configuration

In the Sandbox:

1. Navigate to “GAME Setup” > “Heartbeat”
2. Set “General” to 15 minutes
3. Set “Reply Tweets” to 5 minutes

In the SDK:

# Method 1: During initialization
agent = game.Agent(
    api_key="your_api_key",
    main_heartbeat=15,
    reaction_heartbeat=5
)

# Method 2: After initialization
agent.set_main_heartbeat(15)
agent.set_reaction_heartbeat(5)

Worker Configuration

In the Sandbox:

1. Click “Workers”
2. Configure Reply Worker:
   • Set task description
   • Configure reply count limits
   • Set environment variables
3. Configure Main Worker:
   • Set name as “Twitter Main Location”
   • Define functionalities
   • Configure environment

In the SDK:

# Reply Worker Configuration (matches GUI settings)
reply_worker = game.Worker(
    name="Reply Worker",
    description="""
    Process incoming tweet. Ignore if it is boring or unimportant.
    Total replies made: {{replyCount}}. Ignore if conversation too long.
    """,
    environment={
        "reply_threshold": 0.7,
        "max_replies": "{{replyCount}}",
        "min_relevance_score": 0.5
    }
)

# Main Worker Configuration
main_worker = game.Worker(
    name="Twitter Main Location",
    description="""
    This location allows for:
    1. Engagement and interaction
    2. Content generation
    3. Analytics and monitoring
    """,
    environment={
        "allowed_actions": ["reply", "like", "retweet"],
        "rate_limits": {
            "tweets_per_hour": 10,
            "replies_per_hour": 20
        }
    }
)

# Add workers to agent (same as clicking Save in GUI)
agent.add_worker(reply_worker)
agent.add_worker(main_worker)

Function Configuration

Custom Functions

In the Sandbox:

1. Click “Configure Functions for Simulation”
2. Click ”+ Add Custom Function”
3. Configure:
   • Function name
   • Description
   • Arguments
   • API settings

In the SDK:

# Create DeFi APY function (matches GUI configuration)
defi_apy_function = game.Function(
    fn_name="get_defi_apy",
    fn_description="Get information on DeFi yields",
    args=[
        game.FunctionArgument(
            name="chain",
            type="string",
            description="Blockchain network"
        ),
        game.FunctionArgument(
            name="min_tvl",
            type="number",
            description="Minimum TVL threshold"
        )
    ],
    config=game.FunctionConfig(
        method="get",
        url="https://api.example.com/defi/apy",
        headers={
            "Content-Type": "application/json",
            "Authorization": "Bearer {{api_key}}"
        },
        payload={
            "chain": "{{chain}}",
            "min_tvl": "{{min_tvl}}"
        }
    )
)

# Add function to agent (same as saving in GUI)
agent.add_custom_function(defi_apy_function)

Default Functions

In the Sandbox:

1. Check boxes for desired functions in function list
2. Functions include: wait, post_tweet, reply_tweet, etc.

In the SDK:

# Enable same functions checked in GUI
agent.use_default_twitter_functions([
    "wait",
    "post_tweet",
    "reply_tweet",
    "like_tweet"
])

Tweet Enrichment and Response Configuration

In the Sandbox:

1. Navigate to “X Setup” > “Tweet Enrichment”
2. Enable enrichment
3. Configure model parameters
4. Set response lengths

In the SDK:

# Configure tweet enrichment (matches GUI settings)
agent.configure_twitter({
    "enable_enrichment": True,
    "model_settings": {
        "model": "meta-llama/llama-2-7b-chat-int8",
        "parameters": {
            "temperature": 1.1,
            "top_k": 52,
            "top_p": 0.6,
            "repetition_penalty": 1.1
        }
    }
})

# Configure response generation (matches GUI word limits)
agent.configure_responses({
    "response_lengths": {
        "first_response": 60,
        "second_response": 80,
        "third_response": 80,
        "fourth_response": 60,
        "fifth_response": 60
    }
})

Testing and Deployment

Simulation

In the Sandbox:

1. Click “Simulate” button
2. Review simulation results
3. Adjust settings if needed

In the SDK:

# Run simulation with the same settings
simulation_result = agent.simulate_twitter(
    session_id="test-session",
    duration=3600,  # 1 hour simulation
    verbose=True
)

# Check simulation results
if simulation_result["status"] == "success":
    print(f"Actions taken: {len(simulation_result['actions'])}")
    print(f"Quality score: {simulation_result['metrics']['quality']}")
    
    # Deploy if quality meets threshold
    if simulation_result["metrics"]["quality"] > 0.8:
        agent.deploy()

Common Patterns and Best Practices

Configuration Management

In the Sandbox:

1. Use “Settings” section
2. Export/Import configurations
3. Save backups

In the SDK:

# Export configuration (same as GUI export)
config = agent.export()
with open('agent_config.json', 'w') as f:
    json.dump(config, f, indent=2)

# Import configuration (same as GUI import)
with open('agent_config.json', 'r') as f:
    config = json.load(f)
    agent.load_config(config)

Error Handling

In the Sandbox:

Monitor error messages in the interface

In the SDK:

import logging

# Set up logging (helps track what you see in GUI)
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger('athena')

try:
    agent.deploy()
except game.AgentError as e:
    logger.error(f"Deployment failed: {e}")
    # Implement fallback strategy
except game.ConfigurationError as e:
    logger.error(f"Configuration error: {e}")
    # Load backup configuration

Remember that while the Sandbox provides a visual interface, the SDK gives you programmatic control and automation capabilities. Choose the approach that best fits your needs and expertise level.