OSSS.ai.agents.metadata

OSSS.ai.agents.metadata

Unified Agent Metadata with Multi-Axis Classification.

This module consolidates agent metadata from registry.py and dynamic_composition.py while adding the multi-axis classification system needed for event-driven architecture and future utility agent integration.

AgentMetadata

Bases: BaseModel

Unified agent metadata with multi-axis classification.

Combines functionality from registry-based agent management with dynamic discovery capabilities and adds multi-axis classification for intelligent event routing and service extraction.

validate_string_lists(v) classmethod

Validate string list fields.

validate_agent_class(v) classmethod

Handle agent_class validation and conversion from string.

model_post_init(__context)

Initialize derived fields and defaults.

can_replace(other)

Check if this agent can replace another agent.

Uses both capability matching and version compatibility.

is_compatible_with_task(task_type, domain=None)

Check if this agent is compatible with a specific task type.

Parameters:

Name	Type	Description	Default
task_type	str	Type of task ("transform", "evaluate", "retrieve", etc.)	required
domain	Optional[str]	Optional domain specialization	None

Returns:

Type	Description
bool	True if agent can handle this task type

get_performance_tier()

Get performance tier based on cognitive characteristics.

Returns:

Type	Description
str	Performance tier: "fast", "balanced", "thorough"

to_dict()

Convert to dictionary representation for serialization.

create_for_registry(name, agent_class, requires_llm=False, constructor_pattern=None, description='', dependencies=None, is_critical=True, failure_strategy=FailurePropagationStrategy.FAIL_FAST, fallback_agents=None, health_checks=None, cognitive_speed='adaptive', cognitive_depth='variable', processing_pattern='atomic', primary_capability='', secondary_capabilities=None, pipeline_role='standalone', bounded_context='reflection') classmethod

Create AgentMetadata for registry registration.

This factory method properly handles None values for list fields, letting Pydantic handle default_factory creation instead of manually constructing empty lists.

create_default(name='default_agent', agent_class=None, description='Default agent metadata') classmethod

Create default agent metadata.

from_dict(data) classmethod

Create AgentMetadata from dictionary representation.

TaskClassification

Bases: BaseModel

Classification of work being performed for semantic event routing.

This accepts the raw classifier payload from agent_output_meta["_classifier"] and derives a concrete task_type for NodeExecutionContext / DecisionNode.

Expected upstream payload keys (all optional except task_type when provided): - intent: high-level label, e.g. "action", "question", "query" - confidence: overall intent confidence - domain, domain_confidence - topic, topic_confidence, topics - sub_intent, sub_intent_confidence - labels, raw - model_version - original_text, normalized_text, query_terms

derive_task_type_from_classifier(data) classmethod

Accept raw classifier payloads (like the _classifier block) and derive a concrete task_type if it's missing.

This is what prevents NodeExecutionContext construction from failing when only classifier fields are present.

to_dict()

Convert to dictionary representation - backward compatibility.

from_dict(data) classmethod

Create TaskClassification from dictionary - backward compatibility.

classify_query_task(query)

Classify a user query into task type and characteristics.

This is a simplified rule-based implementation. In production, this could use LLM-powered classification or ML models.