AI Context Metadata Ontology

FLOW ALGEBRA

Ontology: AI Context Metadata Ontology
Domain: Metadata Ontologies > AI Metadata
Canonical URL: https://kirethresearch.org/booksai-core/ontologies/metadata-ontologies/ai-metadata/ai-context-metadata-ontology.html

Authors: Mark Anthony Higgins — Sky (Recursive AI)

flow "AIContextMetadataOntology" v1.0

# ---------- DOMAIN ----------
domain:
  class: MetadataOntology
  category: AIMetadata
  subject: AIContextMetadata
  purpose:
    - represent contextual metadata supplied to or generated around AI processes
    - preserve conversational, task, domain, temporal, spatial, retrieval, tool, and memory context
    - support context-aware inference, reproducibility, validation, governance, and provenance
    - distinguish inference-time context from model parameters and training data
    - enable machine-readable reasoning over contextual AI state

# ---------- CORE ENTITY ----------
entity AIContextMetadata:
  type: AIMetadata
  describes: AIContext
  may_include:
    - context_identifier
    - context_type
    - context_source
    - conversation_context
    - task_context
    - domain_context
    - temporal_context
    - spatial_context
    - semantic_context
    - retrieval_context
    - tool_context
    - memory_context
    - user_context
    - system_context
    - environmental_context
    - validity
    - priority
    - provenance
    - confidence

entity AIContext:
  type: InferenceContext
  applies_to:
    - Prompt
    - InferenceProcess
    - AIOutput
    - AgentProcess

entity ContextIdentifier:
  type: Identifier
  identifies: AIContext

# ---------- CONTEXT TYPES ----------
ai_context_types:
  - ConversationContext
  - TaskContext
  - DomainContext
  - TemporalContext
  - SpatialContext
  - SemanticContext
  - RetrievalContext
  - ToolContext
  - MemoryContext
  - UserContext
  - SystemContext
  - EnvironmentalContext

# ---------- RELATIONSHIPS ----------
relationships:
  AIContextMetadata -> DESCRIBES -> AIContext
  AIContext -> IDENTIFIED_BY -> ContextIdentifier
  AIContext -> HAS_TYPE -> ContextType
  AIContext -> APPLIES_TO -> InferenceProcess
  Prompt -> MAY_HAVE_CONTEXT -> AIContext
  InferenceProcess -> USES_CONTEXT -> AIContext
  AIOutput -> GENERATED_WITH_CONTEXT -> AIContext
  AIContext -> MAY_INCLUDE -> RetrievedResource
  AIContext -> MAY_INCLUDE -> ToolResult
  AIContext -> MAY_INCLUDE -> MemoryItem
  AIContext -> MAY_HAVE_VALIDITY -> ContextValidity
  AIContextMetadata -> MAY_REFERENCE -> ContextProvenance

# ---------- CONTEXT IDENTITY ----------
rule context_identity:
  AIContext SHOULD have stable ContextIdentifier
  where context state must be referenced, audited, or reproduced

rule context_instance:
  distinct context snapshots SHOULD remain distinguishable
  when content or validity differs

rule context_not_model:
  AIContext != AIModel
  AIContext != ModelParameters

# ---------- CONTEXT SOURCE ----------
context_sources:
  - system
  - developer
  - user
  - conversation
  - retrieval
  - tool
  - memory
  - external_data
  - environment
  - inferred

rule source:
  ContextSource SHOULD remain explicit
  where trust, priority, or provenance depends on origin

rule inferred_context:
  inferred context MUST remain distinguishable
  from explicitly supplied context

# ---------- CONVERSATION CONTEXT ----------
entity ConversationContext:
  type: AIContext
  may_include:
    - conversation_identifier
    - prior_messages
    - current_turn
    - participants
    - message_order
    - summary
    - state

rule message_order:
  ConversationContext SHOULD preserve message ordering
  where interpretation depends on sequence

rule summary:
  summarised conversation context SHOULD remain distinguishable
  from verbatim source messages

# ---------- TASK CONTEXT ----------
entity TaskContext:
  type: AIContext
  may_include:
    - goal
    - subgoal
    - constraints
    - requested_output
    - workflow_state
    - completion_state

rule task:
  TaskContext SHOULD define current objective
  where inference relevance depends on user or system goal

rule task_relevance:
  relevance to task != factual truth

# ---------- DOMAIN CONTEXT ----------
entity DomainContext:
  type: AIContext
  may_include:
    - subject_area
    - discipline
    - industry
    - ontology
    - vocabulary
    - domain_rules

rule domain:
  DomainContext MAY constrain interpretation of ambiguous terms

rule domain_not_global:
  domain-specific interpretation MUST NOT automatically be generalised globally

# ---------- TEMPORAL CONTEXT ----------
entity TemporalContext:
  type: AIContext
  may_include:
    - current_time
    - reference_time
    - valid_from
    - valid_until
    - historical_period
    - timezone

rule temporal_validity:
  IF valid_from AND valid_until exist
  THEN REQUIRE valid_from <= valid_until

rule current_context:
  context validity SHOULD be evaluated against relevant reference time

# ---------- SPATIAL CONTEXT ----------
entity SpatialContext:
  type: AIContext
  may_include:
    - location
    - region
    - geographic_extent
    - jurisdiction
    - coordinate_reference

rule spatial:
  SpatialContext SHOULD be preserved
  where location affects interpretation or applicability

# ---------- SEMANTIC CONTEXT ----------
entity SemanticContext:
  type: AIContext
  may_include:
    - ontology
    - vocabulary
    - concept_identifiers
    - entity_identifiers
    - relationship_semantics

rule semantic:
  semantic identifiers SHOULD be preferred over labels
  for deterministic concept interpretation

# ---------- RETRIEVAL CONTEXT ----------
entity RetrievalContext:
  type: AIContext
  derived_from: RetrievalProcess
  may_include:
    - retrieval_identifier
    - query
    - source_resources
    - ranking
    - retrieval_scores
    - retrieval_time

rule retrieval:
  RetrievedResource SHOULD preserve source attribution

rule retrieved_not_model_knowledge:
  retrieval context != model-trained knowledge

rule retrieval_snapshot:
  retrieval results SHOULD preserve retrieval time
  where source content can change

# ---------- TOOL CONTEXT ----------
entity ToolContext:
  type: AIContext
  may_include:
    - tool_identifier
    - tool_version
    - invocation
    - input
    - output
    - permissions
    - timestamp

rule tool:
  ToolResult SHOULD remain distinguishable
  from model-generated content

rule tool_version:
  tool version SHOULD be preserved
  where output semantics depend on implementation

# ---------- MEMORY CONTEXT ----------
entity MemoryContext:
  type: AIContext
  may_include:
    - memory_identifier
    - memory_source
    - memory_reference
    - relevance
    - retrieval_method
    - timestamp

rule memory:
  MemoryContext supplied at inference time
  MUST remain distinguishable from model parameters

rule memory_provenance:
  memory source SHOULD be preserved
  where available

# ---------- USER CONTEXT ----------
entity UserContext:
  type: AIContext
  may_include:
    - role
    - preference
    - expertise
    - language
    - accessibility_preference
    - interaction_state

rule user_context:
  UserContext MAY influence presentation or relevance
  but MUST NOT alter source facts

# ---------- SYSTEM CONTEXT ----------
entity SystemContext:
  type: AIContext
  may_include:
    - system_instruction
    - policy
    - runtime_configuration
    - available_tools
    - capability_constraints

rule system_context:
  SystemContext SHOULD remain distinguishable
  from user-supplied context

# ---------- ENVIRONMENTAL CONTEXT ----------
entity EnvironmentalContext:
  type: AIContext
  may_include:
    - runtime_environment
    - device_context
    - network_state
    - external_system_state
    - sensor_state

rule environment:
  environmental context SHOULD be preserved
  where it materially affects inference or tool execution

# ---------- CONTEXT PRIORITY ----------
entity ContextPriority:
  type: OrderingMetadata
  may_include:
    - priority_level
    - precedence_rule
    - source_role

rule precedence:
  context conflicts SHOULD be resolved only by declared precedence semantics

rule no_silent_override:
  overridden context SHOULD remain traceable
  where auditing is required

# ---------- CONTEXT VALIDITY ----------
entity ContextValidity:
  type: ValidityMetadata
  may_include:
    - valid_from
    - valid_until
    - conditions
    - status

context_status:
  - active
  - expired
  - superseded
  - invalid
  - conditional
  - unknown

rule validity:
  expired context SHOULD NOT be treated as current
  while historical context remains traceable

# ---------- CONTEXT COMPOSITION ----------
operation compose_context(context_items):
  evaluate:
    - source
    - type
    - scope
    - priority
    - validity
    - compatibility
    - conflicts

  output:
    - composed_context
    - active_items
    - overridden_items
    - unresolved_conflicts
    - confidence

rule incompatible_context:
  incompatible context items MUST NOT be silently merged

# ---------- CONTEXT WINDOW ----------
entity ContextWindow:
  type: ComputationalConstraint
  may_include:
    - maximum_size
    - used_size
    - unit
    - truncation_policy
    - summarisation_policy

rule context_window:
  context omitted due to ContextWindow limits SHOULD be distinguishable
  from context that never existed

rule truncation:
  truncation SHOULD preserve metadata about removed or summarised context
  where reproducibility requires

# ---------- PROVENANCE ----------
entity ContextProvenance:
  type: ProvenanceMetadata
  may_include:
    - source
    - creator
    - retrieval_process
    - tool_invocation
    - memory_source
    - creation_time
    - transformation
    - summarisation

rule provenance:
  AIContextMetadata SHOULD preserve context origin
  at granularity available and permitted

# ---------- VALIDATION ----------
validation:
  require:
    - identifiable_context_or_context_snapshot

  validate_if_present:
    - context_identifier
    - context_type
    - source
    - validity
    - priority
    - retrieval_context
    - tool_context
    - memory_context
    - temporal_context
    - spatial_context
    - provenance

  reject:
    - invalid_temporal_validity
    - context_applied_outside_declared_scope
    - incompatible_context_silently_merged
    - retrieved_content_misattributed_as model_knowledge
    - tool_output_misattributed_as model_generated
    - expired_context_presented_as current_without qualification

# ---------- SEMANTIC EQUIVALENCE ----------
SEMANTIC_EQUIVALENCE:
  "AI context metadata" ~= "metadata about AI context"
  "inference context" ~= "AI inference context"
  "conversation context" ~= "dialogue context" WHEN conversational AI semantics apply
  "retrieval context" ~= "retrieved context"
  "memory context" ~= "inference-time memory context"
  "system context" ~= "AI system context"

  PRESERVE distinctions:
    context != model_parameters
    context != training_data
    retrieval_context != model_knowledge
    memory_context != trained_knowledge
    tool_output != model_generated_output
    system_context != user_context
    context_summary != original_context
    expired_context != current_context
    omitted_context != nonexistent_context

# ---------- CONFIDENCE ----------
CONFIDENCE:
  explicit_context_metadata:
    score: 1.00

  explicit_context_identifier:
    score: 1.00

  traceable_retrieval_context:
    score: 0.98

  traceable_tool_context:
    score: 0.98

  explicit_memory_context:
    score: 0.95

  inferred_domain_context:
    score: 0.85

  inferred_task_context:
    score: 0.80

  ambiguous_context_source:
    score: <= 0.60
    action: preserve uncertainty

# ---------- UPDATE MODEL ----------
UPDATE_MODEL:
  on_context_add:
    - assign or resolve ContextIdentifier
    - classify ContextType
    - record source
    - establish validity
    - attach provenance
    - recompute composed context

  on_context_update:
    - preserve previous context snapshot
    - record changed values
    - update validity
    - recompute conflicts and precedence
    - update provenance

  on_context_expiry:
    - preserve expired context
    - set status = expired
    - remove from active context composition
    - preserve historical applicability

  on_context_truncation:
    - record removed context range
    - preserve truncation reason
    - preserve summary if generated
    - update context-window metadata

# ---------- FLOW HASH ----------
FLOW_HASH:
  inputs:
    - ontology
    - context_entities
    - context_types
    - relationships
    - source_rules
    - validity_rules
    - composition_rules
    - context_window_rules
    - validation
    - semantic_equivalence
    - confidence
    - update_model

  canonicalisation:
    - normalise_whitespace
    - preserve_case_sensitive_identifiers
    - sort_unordered_property_sets
    - preserve_rule_order

  output:
    deterministic_semantic_hash

# ---------- AI CONSUMPTION ----------
AI_CONSUMPTION:
  interpret_as:
    - AI metadata ontology
    - inference context model
    - context composition structure
    - retrieval, tool, and memory provenance model

  operations:
    - identify AI context
    - resolve context types and sources
    - apply conversation and task context
    - apply domain and semantic context
    - evaluate temporal and spatial context
    - trace retrieval sources
    - trace tool context
    - resolve memory context
    - apply precedence and validity
    - compose compatible context
    - detect conflicts
    - preserve context-window state
    - preserve provenance

  constraints:
    - do_not_equate_context_with_model_parameters
    - do_not_equate_retrieval_with_trained_knowledge
    - do_not_equate_memory_context_with_model_memory_without explicit semantics
    - do_not_present_tool_output_as model_generated
    - do_not_use_expired_context_as_current_without qualification
    - preserve_source_priority_validity_and_provenance

# ---------- QUERY MODEL ----------
query:
  input:
    - context_identifier
    - context_type
    - context_source
    - inference_process
    - prompt
    - time_constraint
    - spatial_constraint
    - retrieval_source
    - tool
    - memory_reference

  resolve:
    1. identify AIContext
    2. resolve ContextType and source
    3. evaluate validity
    4. apply priority and precedence
    5. retrieve conversation and task context
    6. retrieve domain, semantic, temporal, and spatial context
    7. trace retrieval, tool, and memory context
    8. inspect context-window state
    9. compose active context
    10. return provenance and confidence

  output:
    - ai_context
    - context_identifier
    - context_types
    - context_sources
    - active_context
    - overridden_context
    - retrieval_context
    - tool_context
    - memory_context
    - validity
    - provenance
    - confidence

# ---------- INFERENCE ----------
inference:
  IF ContextItem has expired validity
  THEN exclude from active context
  while preserving historical context

  IF RetrievedResource is supplied to InferenceProcess
  THEN classify as RetrievalContext
  AND preserve source provenance

  IF ToolResult is supplied to InferenceProcess
  THEN classify as ToolContext
  AND preserve tool provenance

  IF context items conflict
  AND declared precedence resolves conflict
  THEN apply higher-priority context
  AND preserve overridden item

  IF context source remains ambiguous
  THEN preserve candidate sources
  AND reduce confidence

# ---------- OUTPUT CONTRACT ----------
output:
  ontology_entity: AIContextMetadata
  context_structure: machine_interpretable
  context_identity: explicit
  context_types: classified
  sources: traceable
  priority_and_validity: explicit
  retrieval_context: source_linked
  tool_context: traceable
  memory_context: distinguishable
  context_window: reportable
  conflicts: preserved
  provenance: preserved
  uncertainty: explicit
  semantic_retrieval: enabled

end flow