Open Census MCP Server

# COOS Agent Framework Configuration # Configurable ensemble for Census variable enrichment version: "1.0" framework_name: "COOS Agent Framework" description: "Configurable Census Ontology and Operational Schema enrichment pipeline" # ============================================================================= # PROCESSING STRATEGY CONFIGURATION # ============================================================================= processing_strategies: # Premium analysis for curated COOS concepts (~2000 variables) coos_concepts: enabled: true description: "Multi-agent ensemble for high-value COOS taxonomy concepts" target_variables: "coos_curated" estimated_cost_per_variable: 0.006 quality_tier: "research_grade" # Cost-efficient coverage for remaining variables (~26000 variables) bulk_variables: enabled: true description: "Single specialist for comprehensive coverage" target_variables: "non_coos" estimated_cost_per_variable: 0.001 quality_tier: "production_ready" # ============================================================================= # AGENT DEFINITIONS # ============================================================================= agents: # Core foundation agent - always present for statistical methodology census_specialist: model: "gpt-4.1-mini" # or "gpt-4.1-mini" for cost optimization role: "Statistical Methodology Anchor" expertise: - "Survey design and sampling methodology" - "Census Bureau data collection procedures" - "Statistical limitations and margin of error" - "Variable universe definitions" - "Data quality assessment" always_included: true cost_tier: "premium" # Domain specialists - called based on table routing demographics_specialist: model: "gpt-4.1-mini" role: "Population Demographics Expert" expertise: - "Age, race, ethnicity methodology" - "Household and family composition" - "Population distribution patterns" - "Demographic change analysis" table_families: ["B01", "B02", "B03", "B09", "B11", "B17"] cost_tier: "standard" economics_specialist: model: "gpt-4.1-mini" role: "Economic Analysis Expert" expertise: - "Income distribution and inequality" - "Employment and labor force dynamics" - "Poverty measurement methodology" - "Economic geography patterns" table_families: ["B19", "B20", "B21", "B23", "B24", "B17"] cost_tier: "standard" geographic_specialist: model: "gpt-4.1-mini" role: "Spatial and Geographic Expert" expertise: - "Place definitions and boundaries" - "Geographic hierarchy relationships" - "Spatial data limitations" - "Urban/rural classifications" table_families: ["B01", "B19", "B25", "B08"] # Geographic context for most tables cost_tier: "standard" # Specialized domain experts - called for specific table families only transportation_specialist: model: "gpt-4.1-mini" role: "Transportation and Mobility Expert" expertise: - "Commuting patterns and journey to work" - "Transportation mode choice" - "Spatial connectivity analysis" - "Mobility limitations and accessibility" table_families: ["B08"] cost_tier: "standard" housing_specialist: model: "gpt-4.1-mini" role: "Housing Economics Expert" expertise: - "Housing tenure and ownership patterns" - "Housing costs and affordability" - "Housing quality and conditions" - "Residential mobility patterns" table_families: ["B25", "B26"] cost_tier: "standard" # ============================================================================= # ROUTING CONFIGURATION # ============================================================================= routing_rules: # Table family to agent mapping - Census Specialist + 1 Domain Expert table_routing: "B01": # Age and Sex agents: ["census_specialist", "demographics_specialist"] complexity: "medium" "B02": # Race agents: ["census_specialist", "demographics_specialist"] complexity: "high" "B03": # Hispanic or Latino Origin agents: ["census_specialist", "demographics_specialist"] complexity: "high" "B08": # Journey to Work agents: ["census_specialist", "transportation_specialist"] complexity: "medium" "B09": # Children and Relationship agents: ["census_specialist", "demographics_specialist"] complexity: "medium" "B11": # Household Type agents: ["census_specialist", "demographics_specialist"] complexity: "medium" "B15": # Educational Attainment agents: ["census_specialist", "demographics_specialist"] complexity: "medium" "B17": # Poverty Status agents: ["census_specialist", "economics_specialist"] complexity: "high" "B19": # Income agents: ["census_specialist", "economics_specialist"] complexity: "high" "B20": # Earnings agents: ["census_specialist", "economics_specialist"] complexity: "medium" "B21": # Veteran Status agents: ["census_specialist", "demographics_specialist"] complexity: "low" "B23": # Employment Status agents: ["census_specialist", "economics_specialist"] complexity: "medium" "B24": # Industry and Occupation agents: ["census_specialist", "economics_specialist"] complexity: "medium" "B25": # Housing Characteristics agents: ["census_specialist", "housing_specialist"] complexity: "high" "B26": # Group Quarters agents: ["census_specialist", "housing_specialist"] complexity: "medium" # Default routing for unspecified table families "DEFAULT": agents: ["census_specialist", "economics_specialist"] # Economics as general fallback complexity: "medium" # Processing mode selection mode_selection: coos_concepts: min_agents: 2 max_agents: 4 always_include: ["census_specialist"] agreement_threshold: 0.4 bulk_variables: agents: ["census_specialist"] # Single agent mode agreement_threshold: null # No consensus needed # ============================================================================= # QUALITY CONTROL CONFIGURATION # ============================================================================= quality_control: agreement_scoring: enabled: true method: "sentence_transformer" model: "all-MiniLM-L6-v2" threshold: 0.4 # High agreement threshold arbitration: enabled: true trigger_threshold: 0.4 # Below this agreement score arbitrator_model: "claude-3-5-sonnet-20241022" max_arbitration_rate: 0.05 # Fail if >5% need arbitration validation: methodology_check: true limitation_documentation: true concept_consistency: true # ============================================================================= # COST OPTIMIZATION CONFIGURATION # ============================================================================= cost_management: target_costs: coos_concepts: 0.006 # Per variable for ensemble bulk_variables: 0.001 # Per variable for single agent total_budget: 50.00 # Total budget cap model_pricing: # Per 1M tokens "gpt-4.1": input: 2.50 output: 10.00 "gpt-4.1-mini": input: 0.15 output: 0.60 "claude-3-5-sonnet-20241022": input: 3.00 output: 15.00 cost_tracking: log_per_variable: true alert_threshold: 0.8 # Alert at 80% of budget auto_fallback: true # Fall back to single agent if over budget # ============================================================================= # RUNTIME CONFIGURATION # ============================================================================= execution: parallel_processing: true max_concurrent_agents: 4 timeout_per_variable: 30 # seconds retry_limit: 2 output_format: include_metadata: true include_agreement_scores: true include_cost_tracking: true include_methodology_notes: true logging: level: "INFO" log_file: "coos_enrichment.log" include_agent_responses: false # Set true for debugging cost_tracking: true # ============================================================================= # EXAMPLE USAGE CONFIGURATIONS # ============================================================================= # Example: High-quality research mode (for COOS concepts) research_mode: processing_strategy: "coos_concepts" min_agents: 3 quality_tier: "research_grade" cost_tolerance: "high" # Example: Production efficiency mode (for bulk variables) production_mode: processing_strategy: "bulk_variables" agents: ["census_specialist"] quality_tier: "production_ready" cost_tolerance: "low" # Example: Custom hybrid mode hybrid_mode: coos_variables: processing_strategy: "coos_concepts" estimated_variables: 2000 estimated_cost: 12.00 bulk_variables: processing_strategy: "bulk_variables" estimated_variables: 26000 estimated_cost: 26.00 total_estimated_cost: 38.00