SUMMARY

• Core API
  • karenina
    • adapters
      • claude_agent_sdk
      • claude_tool
      • factory
      • langchain
      • langchain_deep_agents
      • llm_parallel
      • manual
      • registry
    • benchmark
      • authoring
      • benchmark
      • benchmark_helpers
      • core
      • task_eval
      • verification
    • cli
    • exceptions
    • integrations
    • ports
      • adapter_instruction
      • agent
      • capabilities
      • errors
      • llm
      • messages
      • parser
      • usage
    • scenario
    • schemas
      • checkpoint
      • config
        • models
      • dataframes
        • judgment
        • rubric
        • template
      • entities
        • answer
        • composition
        • question
        • rubric
        • template_spec
        • verified_field
      • outputs
        • rubric
      • primitives
      • results
        • aggregation
        • judgment
        • rubric
        • rubric_judgment
        • template
        • verification_result_set
      • scenario
      • shared
      • verification
        • api_models
        • config
        • config_presets
        • job
        • model_identity
        • prompt_config
        • result
        • result_components
    • storage
    • utils
• Internals
  • karenina
    • adapters
      • claude_agent_sdk
        • agent
        • availability
        • errors
        • llm
        • mcp
        • messages
        • parser
        • prompts
          • deep_judgment
          • parsing
          • rubric
        • registration
        • trace
        • usage
      • claude_tool
        • agent
        • llm
        • mcp
        • messages
        • parser
        • prompts
          • deep_judgment
          • parsing
          • rubric
        • registration
        • tools
        • trace
        • usage
      • langchain
        • agent
        • initialization
        • llm
        • mcp
        • messages
        • middleware
        • models
        • parser
        • prompts
          • deep_judgment
          • feedback_format
          • feedback_null
          • format_instructions
          • parsing
          • rubric
          • summarization
        • registration
        • trace
        • usage
      • langchain_deep_agents
        • agent
        • availability
        • errors
        • initialization
        • llm
        • mcp
        • messages
        • parser
        • prompts
          • deep_judgment
          • parsing
          • rubric
        • registration
        • trace
        • usage
      • manual
        • exceptions
        • helpers
        • manager
        • message_utils
        • registration
        • traces
    • benchmark
      • authoring
        • answers
          • builder
          • generator
          • generator_code
          • generator_prompts
          • reader
        • questions
          • extractor
          • reader
      • core
        • base
        • exports
        • metadata
        • question_query
        • questions
        • results
        • results_io
        • rubrics
        • templates
        • verification_manager
      • task_eval
        • helpers
        • models
        • task_eval
      • verification
        • batch_runner
        • evaluators
          • rubric
            • agentic_trait
            • deep_judgment
            • evaluator
            • llm_trait
            • metric_trait
          • template
            • deep_judgment
            • evaluator
            • results
          • trace
            • abstention
            • sufficiency
        • executor
        • prompts
          • assembler
          • deep_judgment
            • rubric
              • deep_judgment
            • template
              • excerpt
              • hallucination
              • reasoning
          • parsing
            • parsing_instructions
          • rubric
            • literal_trait
            • llm_trait
            • metric_trait
            • presence_check
          • task_types
          • trace
            • abstention
            • sufficiency
        • runner
        • stages
          • core
            • autofail_stage_base
            • base
            • check_stage_base
            • orchestrator
          • helpers
            • deep_judgment_helpers
            • results_exporter
          • pipeline
            • abstention_check
            • agentic_parse_template
            • agentic_rubric_evaluation
            • deep_judgment_autofail
            • deep_judgment_rubric_auto_fail
            • embedding_check
            • finalize_result
            • generate_answer
            • parse_template
            • recursion_limit_autofail
            • rubric_evaluation
            • sufficiency_check
            • trace_validation_autofail
            • validate_template
            • verify_template
        • utils
          • cache_helpers
          • class_discovery
          • embedding_check
          • error_helpers
          • llm_invocation
          • llm_judge_helpers
          • resource_helpers
          • schema_builder
          • search_helpers
          • search_provider
          • search_tavily
          • storage_helpers
          • task_helpers
          • template_converter
          • template_parsing_helpers
          • template_validation
          • trace_agent_metrics
          • trace_formatting
          • trace_fuzzy_match
          • trace_parsing
          • trace_usage_tracker
    • cli
      • interactive
      • optimize
      • preset
      • serve
      • status
      • utils
      • verify
      • verify_config
      • verify_output
    • integrations
      • adele
        • classifier
        • parser
        • prompts
          • system
          • user
        • schemas
        • traits
      • gepa
        • adapter
        • config
        • data_types
        • export
        • feedback
        • prompts
          • defaults
          • feedback
        • scoring
        • splitting
        • tracking
        • verbose
    • scenario
      • builder
      • checkpoint
      • edge_resolution
      • manager
      • outcome_evaluation
      • source_extraction
      • sugar
      • validation
    • schemas
      • primitives
        • comparisons
        • composition
        • normalizers
        • registry
        • scope
        • trace
      • scenario
        • checks
        • definition
        • state
        • types
      • shared
        • search
    • storage
      • auto_mapper
      • base
      • converters
      • db_config
      • engine
      • generated_models
      • migrate_question_keywords
      • migrate_template_id
      • models
      • operations
      • queries
      • rubric_serializer
      • utils
      • views
        • combination_info
        • deep_judgment_rubric_traits
        • models_used
        • question_attributes
        • raw_llm_answers
        • result_mcp_servers
        • result_tools_used
        • results_metadata
        • rubric_traits
        • template_attributes
        • template_results
        • utils
    • utils
      • answer_cache
      • checkpoint
      • checkpoint_trait_converters
      • code
      • errors
      • file_ops
      • json_extraction
      • mcp
        • client
        • tools
      • messages
      • progressive_save
      • retry
      • testing
      • version