​

Memory Manager Agents

​

Available Memory Agents

​

Interpret and Store Memories

​

Quick Start

from erdo.actions import bot

# Store analysis results as memories
result = bot.invoke(
    bot_name="interpret and store memories",
    parameters={
        "code": "df.groupby('category').sum()",
        "output": "Electronics: $50000, Clothing: $30000",
        "context": "Monthly sales analysis by category"
    }
)

​

What Gets Stored

​

Configuration

# Store memories with specific categorization
result = bot.invoke(
    bot_name="interpret and store memories",
    parameters={
        "content": "Customer churn analysis revealed 15% monthly churn rate",
        "category": "business_metrics",
        "context": "Q4 2024 analysis",
        "importance": "high",
        "tags": ["churn", "customers", "retention"]
    }
)

# Store technical insights
result = bot.invoke(
    bot_name="interpret and store memories",
    parameters={
        "code": "pd.read_csv('data.csv', parse_dates=['date'])",
        "insight": "Always parse date columns for time series analysis",
        "category": "technical_tips",
        "applies_to": ["pandas", "data_loading"]
    }
)

​

Search Memories

​

Quick Start

from erdo.actions import bot

# Search for relevant memories
result = bot.invoke(
    bot_name="search memories",
    parameters={
        "query": "How to handle missing customer data?",
        "limit": 5
    }
)

​

Search Capabilities

# Find conceptually related memories
result = bot.invoke(
    bot_name="search memories",
    parameters={
        "query": "sales performance analysis",
        "search_type": "semantic",
        "include_similar": True
    }
)

# Find specific technical solutions
result = bot.invoke(
    bot_name="search memories",
    parameters={
        "query": "pandas groupby error handling",
        "category": "technical_tips",
        "min_relevance": 0.8
    }
)

# Search within specific business context
result = bot.invoke(
    bot_name="search memories",
    parameters={
        "query": "quarterly reporting insights",
        "context": "financial_analysis",
        "time_range": "last_6_months"
    }
)

​

Search Filters

• Category: Filter by memory type (business_metrics, technical_tips, etc.)
• Time Range: Recent memories or historical insights
• Importance Level: High-priority memories first
• Tags: Specific topic areas
• User Context: Personal or team-specific memories

​

Memory Types

​

Business Intelligence Memories

• Key metrics and their definitions
• Recurring analysis patterns
• Decision-making criteria
• Performance benchmarks

​

Technical Memories

• Successful code patterns
• Error resolutions
• Data processing techniques
• Integration solutions

​

Process Memories

• Workflow optimizations
• User preferences
• Reporting templates
• Automation triggers

​

Integration Patterns

# Store insights during analysis
analysis_result = analyze_data(dataset)

# Store the insight for future use
memory_result = bot.invoke(
    bot_name="interpret and store memories",
    parameters={
        "content": analysis_result.insight,
        "context": analysis_result.context,
        "category": "data_insights"
    }
)

# Search for relevant context before analysis
relevant_memories = bot.invoke(
    bot_name="search memories",
    parameters={
        "query": f"analysis of {dataset_name}",
        "limit": 3
    }
)

# Use memories to inform analysis
analysis_context = relevant_memories.memories

# Continuous improvement pattern
def analyze_with_learning(query, dataset):
    # Search for relevant memories
    memories = search_memories(query)

    # Perform analysis with context
    result = analyze_data(dataset, context=memories)

    # Store new insights
    store_memories(result.insights)

    return result

​

Best Practices

​

Memory Organization

• Use consistent categorization schemes
• Include relevant tags for discoverability
• Set appropriate importance levels
• Regular cleanup of outdated memories

​

Search Optimization

• Use specific, descriptive queries
• Leverage category filters for focus
• Combine multiple search approaches
• Validate memory relevance before use

​

Privacy and Security

• Respect data governance policies
• Avoid storing sensitive information
• Implement access controls
• Regular audit of stored memories

​

Advanced Features

# Create connected memories
result = bot.invoke(
    bot_name="interpret and store memories",
    parameters={
        "content": "Sales increased after marketing campaign",
        "related_memories": ["campaign_analysis_123", "sales_metrics_456"],
        "relationship_type": "causal"
    }
)

# Time-aware memory storage
result = bot.invoke(
    bot_name="interpret and store memories",
    parameters={
        "content": "Q4 shows seasonal sales spike",
        "temporal_context": {
            "period": "Q4",
            "year": 2024,
            "seasonality": "high"
        }
    }
)

# Store memories with confidence levels
result = bot.invoke(
    bot_name="interpret and store memories",
    parameters={
        "content": "Customer churn prediction model accuracy: 85%",
        "confidence": 0.85,
        "validation_method": "cross_validation"
    }
)