🔧 Advanced Setup

This guide covers advanced configuration options and production deployment strategies for Ryoma.

🎯 Agent Modes

Ryoma offers different agent modes optimized for various use cases:

ReFoRCE Mode (State-of-the-Art)

agent = SqlAgent(
    model="gpt-4",
    mode="reforce",  # Research-based optimizations
    reforce_config={
        "enable_self_refinement": True,
        "parallel_generation": True,
        "consensus_voting": True,
        "exploration_depth": 3
    }
)

Features:

  • 🚀 Self-refinement workflow

  • 🚀 Format restriction generation

  • 🚀 Column exploration with feedback

  • 🚀 Parallel SQL generation

  • 🚀 Majority-vote consensus

🗄️ Database Configuration

PostgreSQL

from ryoma_data import DataSource

datasource = DataSource(
    "postgres",
    connection_string="postgresql://user:pass@host:5432/db"
)

Snowflake

from ryoma_data import DataSource

datasource = DataSource(
    "snowflake",
    account="your-account",
    user="your-user",
    password="your-password",
    database="your-database",
    warehouse="your-warehouse"
)

BigQuery

from ryoma_data import DataSource

datasource = DataSource(
    "bigquery",
    project_id="your-project",
    credentials_path="/path/to/service-account.json"
)

Database Profiling

from ryoma_data import DataSource, DatabaseProfiler

datasource = DataSource("postgres", connection_string="postgresql://...")

# Configure profiler
profiler = DatabaseProfiler(
    sample_size=10000,
    top_k=10,
    enable_lsh=True
)

# Profile tables
profile = profiler.profile_table(datasource, "customers")

🛡️ Security Configuration

Safety Validation

from ryoma_ai.agent.sql import SqlAgent

agent = SqlAgent(
    model="gpt-4",
    mode="enhanced",
    safety_config={
        "enable_validation": True,
        "allowed_operations": [
            "SELECT", "WITH", "CTE", "UNION", "JOIN"
        ],
        "blocked_operations": [
            "DROP", "DELETE", "UPDATE", "INSERT", "TRUNCATE"
        ],
        "max_rows": 50000,
        "max_execution_time": 300,  # 5 minutes
        "require_where_clause": True,
        "block_cross_database": True
    }
)

Custom Validation Rules

from ryoma_ai.agent.internals.sql_safety_validator import (
    SqlSafetyValidator, ValidationRule
)

# Define custom validation rules
custom_rules = [
    ValidationRule(
        name="no_production_tables",
        pattern=r"FROM\s+prod\.",
        message="Production tables are not allowed",
        severity="error"
    ),
    ValidationRule(
        name="require_limit",
        pattern=r"SELECT.*(?!.*LIMIT)",
        message="All queries must include LIMIT clause",
        severity="warning"
    )
]

# Apply to agent
validator = SqlSafetyValidator(custom_rules=custom_rules)
agent.set_safety_validator(validator)

🔧 Model Configuration

OpenAI Models

agent = SqlAgent(
    model="gpt-4",
    model_parameters={
        "temperature": 0.1,        # Low temperature for consistency
        "max_tokens": 2000,        # Sufficient for complex queries
        "top_p": 0.9,
        "frequency_penalty": 0.1
    }
)

Anthropic Claude

agent = SqlAgent(
    model="claude-3-sonnet-20240229",
    model_parameters={
        "temperature": 0.1,
        "max_tokens": 4000,
        "top_p": 0.9
    }
)

Local Models (GPT4All)

from ryoma_ai.llm.provider import load_model_provider

model = load_model_provider(
    model_id="gpt4all:codellama-13b.gguf",
    model_type="chat",
    model_parameters={"allow_download": True}
)

agent = SqlAgent(
    model=model,
    mode="enhanced"
)

📊 Monitoring and Logging

Enable Detailed Logging

import logging

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
    handlers=[
        logging.FileHandler('ryoma.log'),
        logging.StreamHandler()
    ]
)

# Enable Ryoma-specific logging
ryoma_logger = logging.getLogger('ryoma_ai')
ryoma_logger.setLevel(logging.DEBUG)

Performance Monitoring

from ryoma_ai.monitoring import PerformanceMonitor

# Enable performance tracking
monitor = PerformanceMonitor(
    track_query_time=True,
    track_profiling_time=True,
    track_model_calls=True,
    export_metrics=True
)

agent = SqlAgent(
    model="gpt-4",
    mode="enhanced",
    monitor=monitor
)

🚀 Production Deployment

Docker Configuration

FROM python:3.11-slim

# Install dependencies
COPY requirements.txt .
RUN pip install -r requirements.txt

# Copy application
COPY . /app
WORKDIR /app

# Set environment variables
ENV OPENAI_API_KEY=${OPENAI_API_KEY}
ENV DATABASE_URL=${DATABASE_URL}

# Run application
CMD ["python", "app.py"]

Environment Variables

# API Keys
OPENAI_API_KEY=your-openai-key
ANTHROPIC_API_KEY=your-anthropic-key

# Database Configuration
DATABASE_URL=postgresql://user:pass@host:5432/db
ENABLE_PROFILING=true
PROFILER_SAMPLE_SIZE=10000

# Security Settings
ENABLE_SAFETY_VALIDATION=true
MAX_QUERY_ROWS=50000
ALLOWED_OPERATIONS=SELECT,WITH,CTE

# Performance Settings
AGENT_MODE=enhanced
MODEL_TEMPERATURE=0.1
MAX_RETRIES=3

Health Checks

from ryoma_ai.health import HealthChecker

def health_check():
    checker = HealthChecker()
    
    # Check database connectivity
    db_status = checker.check_database(datasource)
    
    # Check model availability
    model_status = checker.check_model(agent.model)
    
    # Check profiling system
    profiler_status = checker.check_profiler(datasource)
    
    return {
        "database": db_status,
        "model": model_status,
        "profiler": profiler_status,
        "overall": all([db_status, model_status, profiler_status])
    }

🎯 Best Practices

1. Choose the Right Mode

  • Basic: Simple queries, development

  • Enhanced: Production use, complex queries

  • ReFoRCE: Maximum accuracy, research applications

2. Configure Safety Appropriately

  • Always enable validation in production

  • Set appropriate row limits

  • Block dangerous operations

3. Optimize Profiling

  • Use appropriate sample sizes

  • Enable LSH for large schemas

  • Cache profiles for frequently used tables

4. Monitor Performance

  • Track query execution times

  • Monitor model API usage

  • Set up alerts for failures

5. Security Considerations

  • Use environment variables for secrets

  • Implement proper access controls

  • Audit query logs regularly