Crate terraphim_agent_registry 

Terraphim Agent Registry

Knowledge graph-based agent registry for intelligent agent discovery and capability matching.

This crate provides a sophisticated agent registry that leverages Terraphim’s knowledge graph infrastructure to enable intelligent agent discovery, capability matching, and role-based specialization. It integrates with the existing automata and role graph systems to provide context-aware agent management.

Core Features

• Knowledge Graph Integration: Uses existing extract_paragraphs_from_automata and is_all_terms_connected_by_path for intelligent agent discovery
• Role-Based Specialization: Leverages terraphim_rolegraph for agent role management
• Capability Matching: Semantic matching of agent capabilities to task requirements
• Agent Metadata: Rich metadata storage with knowledge graph context
• Dynamic Discovery: Real-time agent discovery based on evolving requirements
• Performance Optimization: Efficient indexing and caching for fast lookups

Re-exports

pub use capabilities::*;

pub use discovery::*;

pub use error::*;

pub use knowledge_graph::*;

pub use matching::*;

pub use metadata::*;

pub use registry::*;

Modules

capabilities

Agent capability management and matching

capability

Capability-based routing types for unified LLM and Agent providers.

discovery

Agent discovery utilities and algorithms

error

Error types for the agent registry

knowledge_graph

Knowledge graph integration for agent registry

llm_usage

LLM usage tracking types for cost monitoring across providers.

matching

Knowledge graph-based agent matching and coordination

mcp_tool

MCP Tool types for indexing and discovery.

metadata

Agent metadata management with role integration

persona

Persona definition types for agent personas with SFIA skill framework support.

procedure

Procedure capture types for the learning system.

registry

Main agent registry implementation

review

Review finding types for multi-agent code review.

score

Scoring algorithms and query types for document relevance ranking. Scoring algorithms and query types for document relevance ranking.

Structs

AgentCommunication

Communication between agents in a multi-agent context

AgentInfo

Information about an AI agent in a multi-agent context

AgentPid

Unique identifier for agents in the supervision system

AgentSpec

Agent specification for creating supervised agents

CapturedProcedure

A captured procedure with ordered steps and execution history.

CharacteristicDef

Definition of a core personality characteristic.

ChatMessage

A single message in a conversation, including metadata for cost tracking.

Concept

A concept is a higher-level, normalized term.

ContextHistory

Context history that tracks what context has been used across conversations

ContextHistoryEntry

Entry in context usage history

ContextItem

Context item that can be added to LLM conversations

Conversation

ConversationId

Unique identifier for conversations

ConversationSummary

Summary of a conversation for listing purposes

CoverageSignal

Coverage governance signal

Document

The central document type representing indexed and searchable content.

Edge

A directed relationship between two nodes in the knowledge graph.

ExtractedEntity

Extracted entity from text

ExtractedRelationship

Extracted relationship from text

GroundingMetadata

Grounding metadata for normalized terms (Dynamic Ontology)

Index

An index is a hashmap of documents

IndexedDocument

Reference to external storage of documents

KGIndexInfo

Knowledge Graph index information

KGTermDefinition

Knowledge Graph term definition with comprehensive metadata

LlmResult

The text content produced by an LLM call together with optional usage metadata.

LlmUsage

Token usage and cost data for a single LLM call.

MarkdownDirectives

Parsed directives extracted from the YAML front matter of a markdown KG entry.

McpToolEntry

Represents an indexed MCP tool from configured servers.

MessageId

Unique identifier for messages within conversations

ModelPricing

Per-model pricing configuration for cost estimation.

MultiAgentContext

Multi-agent context for coordinating between different AI agents

Node

A Node represents single concept and its connections to other concepts.

NormalizedTerm

A normalized term is a higher-level term that has been normalized

NormalizedTermValue

The value of a normalized term

OntologyAntiPattern

Anti-pattern definition for detection

OntologyEntityType

Entity type definition in an ontology schema

OntologyRelationshipType

Relationship type definition in an ontology schema

OntologySchema

Schema-first ontology definition

PatternMatch

Result of pattern matching with priority scoring

PersonaDefinition

A complete persona definition for an AI agent.

Priority

Priority level for routing rules and decisions Higher numeric values indicate higher priority A clamped priority value in the range 0–100 (higher = more urgent).

ProcedureConfidence

Confidence metrics for a procedure based on execution history.

ProcedureStep

A single step in a captured procedure.

ProcessId

Process ID for spawned agents

Provider

A provider that can fulfill capabilities

QualityScore

Quality scores for Knowledge/Learning/Synthesis (K/L/S) dimensions.

ReviewAgentOutput

Output schema for a single review agent’s results.

ReviewFinding

A single structured finding from a review agent.

RoleName

A role name with case-insensitive lookup support.

RouteDirective

Routing directive specifying which LLM provider and model to use.

RoutingDecision

Routing decision with priority information

RoutingRule

A routing rule with pattern matching and priority

SchemaSignal

Schema signal extracted from text

SearchQuery

A search query for finding documents in the knowledge graph.

SfiaSkillDef

SFIA skill definition.

SupervisorId

Unique identifier for supervisors

Thesaurus

A thesaurus is a dictionary with synonyms which map to upper-level concepts.

Enums

Capability

A capability that a provider can fulfill

ContextType

Types of context that can be added to conversations

ContextUsageType

How a context item was used

CostLevel

Cost level for routing decisions

DocumentType

Classifies a document by its role in the knowledge graph.

FindingCategory

Category of a review finding (maps to review groups).

FindingSeverity

Severity level for a review finding.

KnowledgeGraphInputType

Defines all supported inputs for the knowledge graph.

Latency

Latency expectation for routing decisions

Layer

Layered output levels for search results.

LogicalOperator

Logical operators for combining multiple search terms

NormalizationMethod

Normalization method used for grounding

PersonaLoadError

Errors that can occur when loading a persona definition.

ProviderType

Type of provider (LLM or Agent)

RelevanceFunction

Defines the relevance function (scorer) to be used for ranking search results for the Role.

RestartStrategy

Restart strategies for handling agent failures

RotStatus

Health status of conversation context based on token budget utilization.

RoutingScenario

Routing scenario types

Functions

deduplicate_findings

Deduplicate findings by keeping the most severe finding per (file, line, category) key.

extract_first_paragraph

Extract the first paragraph from document body text.

Type Aliases

GenAgentResult

RegistryResult

Result type for agent registry operations