agent-utilities Canonical Concept Registry¶
Note: The machine-generated single source of truth for CONCEPT: tags is now
docs/concepts.yaml(auto-generated byscripts/build_concepts_yaml.pyfromCONCEPT:<ID>markers inagent_utilities/). It currently tracks 70 concepts across 12 pillars. This page is a curated, human-readable narrative map and may lag the generated registry; when the two disagree,concepts.yamlwins.Components marked 🔬 have research-backed enhancements.
Rule: All new concept proposals must go through the DSTDD design phase. See
.specify/design/_template.mdfor the required KG analysis.
Traceability Matrix¶
Every concept has 1:1:1 traceability across:
- Code: CONCEPT:X.Y tag in module docstrings
- Tests: CONCEPT:X.Y tag in test file docstrings
- Docs: Dedicated page in docs/pillars/<pillar>/
Pillar 1: Graph Orchestration Engine (ORCH)¶
| ID | Canonical Name | Code Modules | Tests | Doc Page |
|---|---|---|---|---|
ORCH-1.0 |
Core Orchestration Engine | 17 | 17 | Pillar Summary |
ORCH-1.1 |
HTN Planning Pipeline | 25 | 11 | Pillar Summary |
ORCH-1.2 |
Specialist Routing & Discovery | 28 | 14 | Pillar Summary |
ORCH-1.3 |
Execution Safety & State | 11 | 2 | ORCH-1.3 |
ORCH-1.4 |
Capability Wiring Engine | 11 | 2 | ORCH-1.4 |
ORCH-1.5 |
DSTDD Pipeline | 3 | 3 | ORCH-1.5 |
ORCH-1.6 |
Prediction Linkage Layer 🔬 | 1 | 0 | Pillar Summary |
ORCH-1.7 |
RecursiveMAS Latent Orchestrator 🔬 | 0 | 0 | Pillar Summary |
ORCH-1.8 |
Parallel Execution & Synthesis Engine | 7 | 1 | ORCH-1.8 |
ORCH-1.9 |
Autonomous Department Orchestration | 3 | 1 | Pillar Summary |
ORCH-1.10 |
Reactive Event Sourcing | 3 | 1 | ORCH-1.10 |
ORCH-1.11 |
WASM Micro-Agent Execution | 1 | 1 | OS-5.5 |
ORCH-1.12 |
Structured Predict-RLM Runtime | 2 | 1 | Pillar Summary |
ORCH-1.13 |
GEPA Reflective Prompt Optimizer | 2 | 1 | Pillar Summary |
ORCH-1.27 |
Role-Specialized Model Routing 🔬 | 5 | 2 | ORCH-1.27 |
ORCH-1.28 |
Composable Skills + Generic Adapter 🔬 | 2 | 1 | ORCH-1.28 |
ORCH-1.29 |
RLM Resilience + Telemetry 🔬 | 2 | 1 | ORCH-1.29 |
ORCH-1.30 |
Generalizing GEPA 🔬 | 1 | 1 | ORCH-1.30 |
ORCH-1.31 |
Graph-Native Optimization State 🔬 | 1 | 1 | ORCH-1.31 |
ORCH-1.32 |
KG-Governed Agent Swarm 🔬 | 3 | 1 | ORCH-1.32 |
Key modules: graph/builder.py, graph/nodes.py, graph/planning/ (unified Planner facade), graph/routing/ (Router/RoutingStrategy strategy package over graph/_router_impl.py), graph/executor.py, graph/hsm.py, graph/lifecycle.py, core/default_catalog.py, capabilities/checkpointing.py, sdd/orchestrator.py, graph/kg_graph_factory.py, orchestration/agent_runner.py, graph/parallel_engine.py, graph/manifest_generators.py, models/execution_manifest.py, graph/reactive/ledger.py, graph/reactive/dispatcher.py, core/wasm_runner.py, rlm/predict_rlm.py, rlm/gepa.py, 🔬 graph/coordination.py, 🔬 orchestration/prediction_linkage.py
Pillar 2: Epistemic Knowledge Graph (KG)¶
| ID | Canonical Name | Code Modules | Tests | Doc Page |
|---|---|---|---|---|
KG-2.0 |
Active Knowledge Graph | 42 | 24 | Pillar Summary |
KG-2.1 |
Tiered Memory & Context 🔬 | 17 | 8 | KG-2.1 |
KG-2.2 |
Ontology, Epistemics & DSPy Integration | 27 | 6 | KG-2.2 |
KG-2.3 |
Unified Retrieval & Graph Integrity 🔬 | 15 | 4 | KG-2.3 |
KG-2.4 |
Inductive Knowledge | 12 | 5 | KG-2.4 |
KG-2.5 |
Topological Analysis | 11 | 4 | KG-2.5 |
KG-2.6 |
Domain Ontologies & Vertical Subgraphs | 53 | 34 | KG-2.6 |
KG-2.6 |
Memory Stability | 13 | 2 | Pillar Summary |
KG-2.7 |
Multi-Domain Architecture | 6 | 2 | KG-2.7 |
KG-2.7 |
Centralized Epistemic Gateway & Transaction Proxy | 3 | 3 | KG-2.7 |
KG-2.7 |
Rust-Native High-Performance Compute (FFI) 🔬 | 3 | 3 | OS-5.5 |
KG-2.7 |
Speculative Graph Brancher 🔬 | 1 | 1 | KG-2.7 |
KG-2.7 |
Semantic Compactor & Refactorer 🔬 | 1 | 1 | KG-2.7 |
KG-2.7 |
Single Company Brain | 7 | 2 | Pillar Summary |
KG-2.7 |
Ingestion Engine | 1 | 24 | Pillar Summary |
KG-2.11 |
Bi-Temporal Memory Layers 🔬 | 4 | 1 | KG-2.11 |
KG-2.12 |
Memory-First Retrieval (HyDE) 🔬 | 4 | 1 | KG-2.12 |
KG-2.13 |
Background Learning Engine 🔬 | 2 | 2 | KG-2.13 |
KG-2.14 |
Ground-Truth Context Authority 🔬 | 1 | 1 | KG-2.14 |
KG-2.15 |
Resilient Retrieval 🔬 | 2 | 1 | KG-2.15 |
KG-2.17 |
Memory Hygiene 🔬 | 2 | 1 | KG-2.17 |
KG-2.18 |
Evidence-Weighted Memory 🔬 | 1 | 1 | KG-2.18 |
KG-2.19 |
Self-Curating Wiki 🔬 | 2 | 1 | KG-2.19 |
KG-2.20 |
Mementified Context Management 🔬 | 4 | 1 | KG-2.20 |
KG-2.21 |
Working Set Eviction & Memory Management | 1 | 1 | KG-2.21 |
KG-2.22 |
Pack-Driven Retrieval Signals 🔬 | 3 | 2 | KG-2.22 |
KG-2.23 |
Rust-Accelerated Reasoning | 1 | 1 | KG-2.23 |
KG-2.24 |
Live Refreshable Artifact models + bounded-JSON + safe interpolation | 1 | 1 | KG-2.24 |
KG-2.25 |
the default registry is populated at import, not an empty | 1 | 1 | KG-2.25 |
KG-2.26 |
Trade-Journal Bias Auditor and Shadow Account | 1 | 1 | KG-2.26 |
KG-2.27 |
Agent Calibration and Reputation Tracking | 1 | 1 | KG-2.27 |
KG-2.28 |
Persona Decision-Heuristic Enrichment | 1 | 1 | KG-2.28 |
KG-2.29 |
Sentiment Fusion Signals | 1 | 1 | KG-2.29 |
KG-2.30 |
Geopolitical Risk Scoring | 1 | 1 | KG-2.30 |
KG-2.31 |
Dividend Sustainability & Credit/Fixed-Income Quality | 1 | 1 | KG-2.31 |
KG-2.32 |
Multi-Market Composite Backtester | 1 | 1 | KG-2.32 |
KG-2.33 |
Zero-LLM Pack-Driven Link Inference 🔬 | 2 | 1 | KG-2.33 |
KG-2.34 |
Relational-Intent Retrieval 🔬 | 2 | 1 | KG-2.34 |
KG-2.35 |
Schema-Pack Lifecycle and Audit 🔬 | 3 | 2 | KG-2.35 |
KG-2.36 |
Pack-Driven OWL Closure 🔬 | 2 | 1 | KG-2.36 |
KG-2.37 |
Research-State Domain Pack 🔬 | 2 | 1 | KG-2.37 |
Schema-Pack 2.0 (KG-2.22–KG-2.37) turns the domain Schema Pack from a type-selection profile into a full domain retrieval+extraction+reasoning profile, closing gbrain-class gaps (recency decay, source-trust, zero-LLM typed-edge extraction, relational recall, autocut, candidate auditing) while leveraging our OWL reasoner and bi-temporal store for closure and
as_ofliterature state that a flat brain layer cannot provide. Seemodels/schema_pack.py,models/schema_pack_loader.py,models/schema_pack_audit.py,knowledge_graph/retrieval/{autocut,relational_intent}.py,knowledge_graph/kb/link_inference.py, and theresearch-statepack.
Key modules: knowledge_graph/core/engine.py, knowledge_graph/core/engine_memory.py, knowledge_graph/core/engine_tasks.py, knowledge_graph/core/graph_compute.py, knowledge_graph/core/topological_analysis_engine.py, knowledge_graph/research/research_intelligence_engine.py, knowledge_graph/memory/synthesis.py, knowledge_graph/memory/memory_materializer.py, knowledge_graph/memory/observer.py, knowledge_graph/memory/reflector.py, knowledge_graph/memory/startup_context.py, knowledge_graph/ontology.ttl, knowledge_graph/retrieval/retrieval_quality.py, knowledge_graph/pipeline/document_deletion.py, knowledge_graph/pipeline/document_update.py, domains/finance/, knowledge_graph/orchestration/engine_enterprise.py, knowledge_graph/pipeline/phases/external_graphs.py, knowledge_graph/ingestion/engine.py, scripts/install_git_hooks.py, scripts/submit_diff.py, mcp/kg_server.py, mcp/kg_coordinator.py, knowledge_graph/backends/epistemic_graph_backend.py, knowledge_graph/backends/postgresql_backend.py, knowledge_graph/backends/tiered_backend.py, knowledge_graph/backends/contrib/ladybug_backend.py, 🔬 knowledge_graph/core/ar_graph.py, 🔬 knowledge_graph/core/time_series_graph.py
Pillar 3: Agentic Harness Engineering (AHE)¶
| ID | Canonical Name | Code Modules | Tests | Doc Page |
|---|---|---|---|---|
AHE-3.0 |
Agentic Harness Core | 16 | 3 | Pillar Summary |
AHE-3.1 |
Continuous Evaluation & DSPy Math Optimization | 17 | 7 | AHE-3.1 |
AHE-3.2 |
Agentic Evolution Engine | 16 | 5 | AHE-3.2 |
AHE-3.3 |
Team & Synergy Optimization | 14 | 5 | Pillar Summary |
AHE-3.4 |
Distributed Agentic Evolution | 11 | 1 | AHE-3.4 |
AHE-3.5 |
Heavy Thinking & Background Intelligence | 11 | 1 | AHE-3.5 |
AHE-3.6 |
Backtest & Curriculum | 10 | 2 | AHE-3.6 |
AHE-3.7 |
KG-Native Task Detection 🔬 | 1 | 0 | Pillar Summary |
AHE-3.8 |
Interpretability & Model Evolution | 4 | 2 | Pillar Summary |
AHE-3.9 |
Physical Knowledge Distillation Engine 🔬 | 1 | 1 | Pillar Summary |
AHE-3.10 |
Multi-Optimizer Prompt Selection Strategy 🔬 | 1 | 1 | Pillar Summary |
AHE-3.11 |
GitOps Commit & Evolution Boundary Traceability 🔬 | 1 | 1 | Pillar Summary |
AHE-3.12 |
LongMemEval-S Validation Harness 🔬 | 2 | 1 | AHE-3.12 |
Key modules: harness/evaluation_engine.py, harness/agentic_evolution_engine.py, graph/team_composer.py, agentic_evolution/forge.py, knowledge_graph/orchestration/engine_ahe.py, knowledge_graph/distillation/physical_distiller.py, harness/evolve_agent.py, 🔬 harness/distributed_state_manager.py
Pillar 4: Ecosystem & Peripherals (ECO)¶
| ID | Canonical Name | Code Modules | Tests | Doc Page |
|---|---|---|---|---|
ECO-4.0 |
Tool Interface & MCP Factory | 26 | 18 | ECO-4.0 |
ECO-4.1 |
A2A Network & Consensus 🔬 | 7 | 3 | Pillar Summary |
ECO-4.2 |
Community Telemetry & Ecosystem Map | 5 | 1 | ECO-4.2 |
ECO-4.3 |
Market Data KG Node Models | 1 | 1 | Pillar Summary |
ECO-4.4 |
KG MCP Server & Execution | 2 | 0 | Pillar Summary |
ECO-4.5 |
Native Messaging Backend Abstraction | 21 | 17 | ECO-4.5 |
ECO-4.6 |
Dynamic Capability Ingestion & Discovery | 5 | 4 | ECO-4.6 |
ECO-4.7 |
Company Infrastructure Orchestration | 3 | 2 | Pillar Summary |
ECO-4.8 |
Infrastructure Blueprint Library | 1 | 1 | Pillar Summary |
ECO-4.9 |
Pluggable Event Queue Backend | 3 | 2 | OS-5.5 |
ECO-4.10 |
Automated Documentation & AGENTS.md Governance | 4 | 0 | Pillar Summary |
ECO-4.11 |
Deterministic Lint Enforcement Hook | 1 | 0 | Pillar Summary |
ECO-4.12 |
Plugin Bundle Distribution System | 1 | 0 | Pillar Summary |
ECO-4.13 |
Ecosystem Governance & Policy Engine | 3 | 0 | Pillar Summary |
Key modules: mcp/server_factory.py, mcp/kg_server.py (incl. kg_launch_terminal_agent), ecosystem/bridge.py, ecosystem/hook_installer.py, ecosystem/agents_md_reflector.py, ecosystem/lint_enforcement_hook.py, ecosystem/plugin_bundle.py, ecosystem/permission_policy.py, ecosystem/config_staleness_auditor.py, ecosystem/governance_workflow.py, ecosystem/agent_manager_dashboard.py, tools/codebase_map_tools.py, knowledge_graph/core/agents_md.py, knowledge_graph/memory/startup_context.py, graph/subagent_patterns.py, protocols/a2a_graph_skill.py, tools/tool_filtering.py, tools/dynamic_tool_orchestrator.py, knowledge_graph/core/engine_ingestion.py, knowledge_graph/core/engine_mcp_discovery.py, knowledge_graph/core/queue_backend.py, knowledge_graph/core/nats_queue_backend.py, knowledge_graph/core/kafka_queue_backend.py
Pillar 5: Agent OS Infrastructure (OS)¶
| ID | Canonical Name | Code Modules | Tests | Doc Page |
|---|---|---|---|---|
OS-5.0 |
Agent OS Kernel & XDG Paths | 17 | 6 | Pillar Summary |
OS-5.1 |
Security & Auth | 19 | 8 | OS-5.1 |
OS-5.2 |
Resource Scheduling 🔬 | 18 | 4 | Pillar Summary |
OS-5.3 |
OS Guardrails & Safety Boundaries | 10 | 5 | OS-5.3 |
OS-5.4 |
Telemetry & Observability | 7 | 3 | OS-5.4 |
OS-5.5 |
Massive Scale Architecture & Sandbox | 2 | 2 | OS-5.5 |
OS-5.6 |
Distributed Replay & Compliance Engine | 1 | 1 | OS-5.6 |
OS-5.7 |
OS-Level Hardened Tool Sandbox Executor | 1 | 1 | OS-5.7 |
OS-5.8 |
Epistemic Resource Scheduler 🔬 | 1 | 1 | OS-5.8 |
Key modules: core/paths.py, security/guardrails.py, security/tool_guard.py, core/cognitive_scheduler.py, observability/token_tracker.py, observability/audit_logger.py, graph/reactive/budget.py, core/wasm_runner.py, gateway/aggregator.py, gateway/registry.py, gateway/config.py, gateway/api.py, gateway/ws.py
Gateway Service Dashboard (GW)¶
| ID | Canonical Name | Code Modules | Tests | Doc Page |
|---|---|---|---|---|
OS-5.9 |
Gateway Service Dashboard | 58 | 7 | OS-5.9 |
Key modules: gateway/__init__.py, gateway/models.py, gateway/registry.py, gateway/config.py, gateway/aggregator.py, gateway/api.py, gateway/ws.py, gateway/widgets/base.py, gateway/widgets/*.py (50 widget modules)
Concept Lifecycle¶
New Feature Request
│
▼
┌────────────────────┐
│ KG Analogy Search │ ← Does a similar concept already exist?
│ (similarity ≥ 0.7) │
└────────┬───────────┘
│
┌──────┴──────┐
│ │
EXTEND PROPOSE
│ │
▼ ▼
Augment NewConceptProposal
existing (requires C4 diagram,
concept pillar assignment,
pipeline phase)
│ │
└──────┬──────┘
│
▼
.specify/design/<feature>/design.md
│
▼
SDDManager.validate_design()
│
▼
.specify/specs/<feature>/spec.md
Pillar 6: GeniusBot Desktop Cockpit (GBOT)¶
| ID | Canonical Name | Code Modules | Tests | Doc Page |
|---|---|---|---|---|
GBOT-6.0 |
Desktop Cockpit Orchestrator | geniusbot/geniusbot.py |
tests/test_geniusbot.py |
GeniusBot Cockpit |
GBOT-6.1 |
Ecosystem Dynamic Tab Matrix | geniusbot/plugins/ |
tests/test_plugins.py |
GeniusBot Cockpit |
GBOT-6.2 |
Embedded Terminal Sandbox | geniusbot/qt/terminal_widget.py |
tests/test_terminal_widget.py |
GeniusBot Cockpit |
GBOT-6.3 |
Universal Tool Approval Gate | geniusbot/qt/tool_guard.py |
tests/test_tool_guard.py |
GeniusBot Cockpit |
GBOT-6.4 |
Topological Cockpit Memory | geniusbot/utils/agent_bridge.py |
tests/test_agent_bridge.py |
GeniusBot Cockpit |
GBOT-6.5 |
Multi-Tenant Daemon & Tray | geniusbot/utils/daemon.py |
tests/test_daemon.py |
GeniusBot Cockpit |
GBOT-6.6 |
High-Performance Visual Finance Cockpit | geniusbot/qt/finance_cockpit.py |
tests/test_finance_cockpit.py |
GeniusBot Cockpit |
Cross-Repo: LLM Trainer (CONCEPT:ML-*)¶
A deliberate cross-repo concept family (spans
data-science-mcp+agent-utilities+universal-skills), so it uses a repo-neutralML-*prefix rather than a single pillar. It expandsAHE-3.1(in-house training substrate) andDSCI-004. Not part of the per-pillar contiguity gate. The agent-utilities side is prompt personas + thetrain_modelworkflow (no in-codeCONCEPT:markers, so it does not appear inconcepts.yaml). Canonical registry:data-science-mcp/docs/concepts.md; design:.specify/specs/llm-model-trainer/; deep dive:architecture/in_house_training_substrate.md.
| ID | Canonical Name | Home repo |
|---|---|---|
ML-001 |
Trainer Hardening (shared run_loop) | data-science-mcp |
ML-002 |
Corpus Curation Engine | data-science-mcp |
ML-003 |
Pretrain From Random Init | data-science-mcp |
ML-004 |
Experiment Tracking (MLflow + KG mirror) | data-science-mcp |
ML-005 |
Distributed Scale-Out (FSDP + DeepSpeed) | data-science-mcp |
ML-006 |
Benchmark Evaluation (lm-eval) | data-science-mcp |
ML-007 |
Agent-Driven Training (personas + train_model workflow) |
agent-utilities + universal-skills |
Concept ID Completeness (registry reconciliation)¶
This section keeps the curated narrative map's ID space contiguous (the
check-concept-gapsgovernance gate requires0..maxwith no holes). It reconciles two cases that previously left gaps: (a) concepts that exist indocs/concepts.yaml/ code but were never written into the narrative tables above, and (b) ID numbers that were never assigned (skipped during early numbering) — documented here as Reserved so the number is accounted for rather than silently missing.docs/concepts.yamlremains the machine source of truth.
Pillar 1 — Graph Orchestration (ORCH) — reconciled IDs¶
| ID | Canonical Name | Status |
|---|---|---|
ORCH-1.14 |
— | Reserved (never assigned) |
ORCH-1.15 |
— | Reserved (never assigned) |
ORCH-1.16 |
— | Reserved (never assigned) |
ORCH-1.17 |
— | Reserved (never assigned) |
ORCH-1.18 |
— | Reserved (never assigned) |
ORCH-1.19 |
— | Reserved (never assigned) |
ORCH-1.20 |
Service Registry Initialization | Active |
ORCH-1.21 |
Execution Provenance Tracking | Active |
ORCH-1.22 |
Workflow Persistence & Replay | Active |
ORCH-1.23 |
Semantic Workflow Retrieval | Active |
ORCH-1.24 |
Workflow Lifecycle Management | Active |
ORCH-1.25 |
— | Reserved (never assigned) |
ORCH-1.26 |
RLM Synthesis Fallback | Active |
Pillar 2 — Epistemic Knowledge Graph (KG) — reconciled IDs¶
| ID | Canonical Name | Status |
|---|---|---|
KG-2.8 |
Ingestion & Enrichment Engine | Active |
KG-2.9 |
Universal DataConnector | Active |
KG-2.10 |
Orchestration Synthesis | Active |
KG-2.16 |
Memory Consolidation Stability | Active |