RWKV Agent Kit provides a flexible configuration system that supports multiple configuration methods and rich configuration options. This guide will detail how to configure and optimize your intelligent agents.

📋 Configuration File Formats

RWKV Agent Kit supports TOML, JSON, and YAML format configuration files:

TOML Format (Recommended)

# config.toml
[model]
path = "./models/rwkv-4-world-7b-v1-20230626-ctx4096.pth"
device = "wgpu"  # or "cpu"
precision = "fp16"  # or "fp32"
max_tokens = 2048
temperature = 0.7
top_p = 0.9
top_k = 40

[memory]
type = "sqlite"
path = "./data/agent_memory.db"
max_context_length = 4096
memory_threshold = 0.8
embedding_model = "sentence-transformers/all-MiniLM-L6-v2"

[database]
url = "sqlite://./data/agent.db"
max_connections = 10
connection_timeout = 30

[logging]
level = "info"
file = "./logs/agent.log"
max_file_size = "10MB"
max_files = 5

[tools]
enable_builtin = true
custom_tools_dir = "./tools"

[performance]
thread_pool_size = 4
batch_size = 8
cache_size = 1000

JSON Format

{
  "model": {
    "path": "./models/rwkv-model.pth",
    "device": "wgpu",
    "precision": "fp16",
    "max_tokens": 2048,
    "temperature": 0.7,
    "top_p": 0.9,
    "top_k": 40
  },
  "memory": {
    "type": "sqlite",
    "path": "./data/memory.db",
    "max_context_length": 4096,
    "memory_threshold": 0.8
  },
  "database": {
    "url": "sqlite://./data/agent.db"
  },
  "logging": {
    "level": "info",
    "file": "./logs/agent.log"
  }
}

🎛️ Configuration Options Details

Model Configuration (model)

Option	Type	Default	Description
path	String	Required	RWKV model file path
device	String	"cpu"	Runtime device: "cpu", "wgpu", "mps"
precision	String	"fp32"	Precision: "fp16", "fp32"
max_tokens	u32	2048	Maximum generation tokens
temperature	f32	0.7	Temperature parameter (0.0-2.0)
top_p	f32	0.9	Top-p sampling parameter
top_k	u32	40	Top-k sampling parameter
repetition_penalty	f32	1.1	Repetition penalty

Memory Configuration (memory)

Option	Type	Default	Description
type	String	"sqlite"	Storage type: "sqlite", "postgresql"
path	String	"./memory.db"	SQLite database path
url	String	-	PostgreSQL connection URL
max_context_length	u32	2048	Maximum context length
memory_threshold	f32	0.7	Memory storage threshold
embedding_model	String	"all-MiniLM-L6-v2"	Embedding model name
max_memories	u32	10000	Maximum number of memories

Database Configuration (database)

Option	Type	Default	Description
url	String	Required	Database connection URL
max_connections	u32	10	Maximum connections
connection_timeout	u64	30	Connection timeout (seconds)
idle_timeout	u64	600	Idle timeout (seconds)
enable_logging	bool	false	Enable SQL logging

Logging Configuration (logging)

Option	Type	Default	Description
level	String	"info"	Log level: "trace", "debug", "info", "warn", "error"
file	String	-	Log file path
max_file_size	String	"10MB"	Maximum file size
max_files	u32	5	Maximum number of files
format	String	"json"	Log format: "json", "text"

🚀 Performance Optimization Configuration

GPU Acceleration Configuration

[model]
device = "wgpu"
precision = "fp16"  # Use half precision for speed
batch_size = 8      # Batch size

[performance]
gpu_memory_fraction = 0.8  # GPU memory usage ratio
enable_tensor_cores = true # Enable Tensor Cores

Memory Optimization Configuration

[memory]
max_context_length = 1024  # Reduce context length
memory_threshold = 0.9     # Increase memory threshold
max_memories = 5000        # Limit number of memories

[performance]
cache_size = 500           # Reduce cache size
thread_pool_size = 2       # Reduce thread count

High Concurrency Configuration

[database]
max_connections = 50
connection_timeout = 10

[performance]
thread_pool_size = 16
batch_size = 16
max_concurrent_requests = 100

🔧 Environment Variable Configuration

You can use environment variables to override settings in configuration files:

# Model configuration
export RWKV_MODEL_PATH="/path/to/model.pth"
export RWKV_DEVICE="wgpu"
export RWKV_PRECISION="fp16"

# Database configuration
export RWKV_DATABASE_URL="sqlite://./data/agent.db"

# Logging configuration
export RWKV_LOG_LEVEL="debug"
export RWKV_LOG_FILE="./logs/debug.log"

Usage in code:

use rwkv_agent_kit::config::AgentConfig;

// Automatically load configuration from environment variables
let config = AgentConfig::from_env()?;

// Or merge configuration file and environment variables
let config = AgentConfig::from_file("config.toml")?
    .merge_env()?;

📊 Configuration Validation

Validate Configuration File

use rwkv_agent_kit::config::{AgentConfig, ConfigError};

match AgentConfig::from_file("config.toml") {
    Ok(config) => {
        // Validate configuration
        if let Err(e) = config.validate() {
            eprintln!("Configuration validation failed: {}", e);
            return;
        }
        println!("Configuration validation successful");
    }
    Err(ConfigError::FileNotFound) => {
        eprintln!("Configuration file not found");
    }
    Err(ConfigError::ParseError(e)) => {
        eprintln!("Configuration file parse error: {}", e);
    }
    Err(e) => {
        eprintln!("Other configuration error: {}", e);
    }
}

Configuration Check Tool

# Use command line tool to check configuration
rwkv-agent-kit check-config config.toml

# Example output
✓ Model file exists
✓ Database connection normal
✓ Memory configuration reasonable
⚠ GPU unavailable, will use CPU
✓ Configuration validation passed

🎯 Use Case Configurations

Development Environment

[model]
device = "cpu"
max_tokens = 512

[logging]
level = "debug"
file = "./logs/dev.log"

[memory]
max_memories = 1000

Production Environment

[model]
device = "cuda"
precision = "fp16"
max_tokens = 2048

[logging]
level = "warn"
file = "/var/log/rwkv-agent/agent.log"
max_file_size = "100MB"

[database]
url = "postgresql://user:pass@localhost/rwkv_agent"
max_connections = 50

[performance]
thread_pool_size = 16
cache_size = 5000

Edge Devices

[model]
device = "cpu"
precision = "fp32"
max_tokens = 256

[memory]
type = "sqlite"
max_context_length = 512
max_memories = 500

[performance]
thread_pool_size = 2
batch_size = 1
cache_size = 100

🔍 Troubleshooting

Common Configuration Issues

1. Model Loading Failed

   Error: Unable to load model file
   Solution: Check if model path is correct and file exists

2. GPU Memory Insufficient

   Error: CUDA out of memory
   Solution: Reduce batch_size or use fp16 precision

3. Database Connection Failed

   Error: Unable to connect to database
   Solution: Check database URL and permission settings

Performance Tuning Recommendations

1. CPU Optimization

   • Set appropriate thread_pool_size
   • Use fp32 precision
   • Reduce batch_size

2. GPU Optimization

   • Use fp16 precision
   • Increase batch_size
   • Enable tensor_cores

3. Memory Optimization

   • Reduce max_context_length
   • Limit max_memories
   • Adjust cache_size

---

Need more help? Check API Documentation or visit GitHub Issues.