Skip to main content

Organizational Context Continuity

SOMA's organizational context continuity system enables seamless knowledge sharing and context preservation across different operators, sessions, and Claude Code instances (CLI, desktop, web, VS Code).

Overview​

Traditional AI agent systems operate in isolation, losing valuable organizational context when switching between different interfaces or operators. SOMA's organizational context system addresses this by:

  • Cross-operator intelligence - Knowledge and patterns identified by one operator benefit the entire organization
  • Multi-instance continuity - Context flows seamlessly between CLI, desktop, web, and IDE instances
  • Team-scoped working memory - L2 working memory respects team boundaries while enabling collaboration
  • Session correlation - Related work sessions are automatically linked across time and operators

Core Components​

Operator Identity Tracking​

Every action in SOMA includes operator context:

interface OperatorContext {
  operatorId: string;
  sessionId: string;
  teamId?: string;
  instanceId?: string; // CLI, desktop, web, vscode
}

This enables:

  • Attribution of insights and patterns to specific operators
  • Cross-operator validation of organizational knowledge
  • Team-based access control and privacy boundaries

Session Correlation​

SOMA automatically correlates related sessions using multiple strategies:

  1. Continuity markers - Explicit handoffs between sessions
  2. Temporal correlation - Sessions within configurable time windows
  3. Problem similarity - Sessions working on related challenges
  4. Multi-instance correlation - Same operator across different Claude Code types

Cross-Operator Pattern Detection​

Patterns identified by multiple operators gain higher confidence scores through:

  • Operator consistency analysis - How often different operators observe the same patterns
  • Multi-operator validation - Formal validation workflows for critical patterns
  • Consensus mechanisms - Automated and manual consensus building for pattern promotion

Governance and Quality Control​

Multi-Operator Validation​

Critical organizational knowledge requires validation from multiple operators:

interface ValidationRequest {
  entryId: string;
  validationType: 'promotion' | 'accuracy' | 'relevance';
  requiredValidators: number;
  deadline: number;
}

Confidence Thresholds​

Auto-promotion thresholds vary by pattern type and evidence quality:

  • Workflow patterns - Standard confidence requirements
  • Anti-patterns - Higher validation bar due to risk
  • Cross-operator patterns - Elevated due to organizational impact
  • Tool usage patterns - Context-dependent based on tool criticality

Audit Logging​

All organizational policy changes are logged with:

  • Impact assessment - Scope (individual/team/organization/system)
  • Risk evaluation - Reversibility and potential consequences
  • Compliance tracking - Approval requirements and governance workflows
  • Operator accountability - Full operator context and session details

Team-Scoped Working Memory​

Privacy Boundaries​

L2 working memory respects organizational structure:

  • Team isolation - Teams cannot access other teams' working memory
  • Cross-team patterns - Elevated to L3 for broader organizational benefit
  • Access control - Policy bridge enforces team membership validation

Decay Windows​

Different teams can configure different decay policies:

const teamDecayConfig = {
  'engineering': 7, // 7 days
  'design': 14,     // 14 days
  'research': 30    // 30 days
};

Claude Code Integration​

Multi-Instance Correlation​

SOMA tracks and correlates activity across all Claude Code interfaces:

  • CLI sessions - Command-line tool usage and automation patterns
  • Desktop app - Interactive conversation and file manipulation
  • Web interface - Browser-based interactions and sharing
  • IDE extensions - Code-focused patterns and development workflows

Workflow Pattern Recognition​

Identifies common organizational workflows:

  • Development patterns - Code review, debugging, refactoring workflows
  • Documentation patterns - Knowledge capture and sharing practices
  • Infrastructure patterns - Deployment, monitoring, and maintenance workflows
  • Research patterns - Investigation and analysis methodologies

Anti-Pattern Detection​

Automatically identifies inefficient practices:

  • Tool redundancy - Unnecessary tool switching or duplication
  • Context loss - Failed handoffs between sessions or operators
  • Knowledge silos - Information trapped in individual contexts
  • Process inefficiencies - Suboptimal workflows or decision patterns

Performance Optimization​

Materialized Indexes​

Efficient queries for organizational context:

// O(1) operator-scoped queries
vault.listByOperator('operator-123', { layer: 'working' });

// O(1) team-scoped queries  
vault.listByTeam('engineering', { type: 'decision' });

Session-Aware Batching​

Reduces I/O overhead for related operations:

  • Time-based batching - Group operations within configurable windows
  • Session affinity - Prefer batching operations from the same session
  • Atomic execution - Option for all-or-nothing batch processing

Best Practices​

For Operators​

  1. Use descriptive session names - Helps with correlation and discovery
  2. Tag organizational insights - Use consistent tagging for better pattern detection
  3. Participate in validation - Review and validate patterns from other operators
  4. Respect team boundaries - Don't attempt to access cross-team working memory

For Teams​

  1. Configure appropriate decay windows - Balance memory efficiency with context retention
  2. Establish validation workflows - Define processes for critical knowledge validation
  3. Review auto-promoted patterns - Regular audits of automatically promoted organizational knowledge
  4. Monitor cross-team patterns - Watch for opportunities to break down silos

For Organizations​

  1. Define governance policies - Clear rules for knowledge validation and promotion
  2. Monitor audit logs - Regular review of organizational policy changes
  3. Tune confidence thresholds - Adjust based on organizational risk tolerance
  4. Analyze operator patterns - Identify training opportunities and process improvements

Configuration​

See Configuration Reference for detailed configuration options including:

  • Team decay window configuration
  • Confidence threshold tuning
  • Session correlation parameters
  • Audit logging and compliance settings
  • Performance optimization options

Security Considerations​

Organizational context continuity involves sharing information across operators and sessions. Key security measures include:

  • Data masking - Automatic masking of sensitive information in audit logs
  • Access controls - Strict enforcement of team membership and permissions
  • Audit trails - Complete logging of all organizational knowledge changes
  • Privacy boundaries - Technical enforcement of team isolation in L2 working memory