Skip to content

Heavy Thinking Orchestration (CONCEPT:AHE-3.4)

Overview

Two-stage parallel-then-deliberate reasoning pipeline with tiered complexity gating, trajectory pruning/shuffling, iterative refinement, and KG-native trajectory persistence (CONCEPT:AHE-3.4). The former HeavyThinkingOrchestrator is now consolidated into the unified ParallelEngine (same code path for a 1-agent call up to a 300-agent swarm).

Implementation Details

  • Source Code: agent_utilities/graph/parallel_engine.py
  • Pillar: AHE

Documentation Coverage

This is an auto-generated dedicated concept page to ensure 100% documentation coverage across the ecosystem.

Continual Learning & Experience (CONCEPT:AHE-3.4)

Overview

Experience distillation via cross-rollout critique (CONCEPT:AHE-3.4), decomposed context retrieval (CONCEPT:AHE-3.4), and Memory-Aware Test-Time Scaling (CONCEPT:AHE-3.4).

Implementation Details

  • Source Code: agent_utilities/graph/verification.py
  • Pillar: AHE

Documentation Coverage

This is an auto-generated dedicated concept page to ensure 100% documentation coverage across the ecosystem.

KG-Native Agentic Task Detection (CONCEPT:AHE-3.4)

Overview

Evaluates topological complexity to route dense API toolchains to complex models.

Implementation Details

  • Source Code: agent_utilities/graph/_router_impl.py
  • Pillar: AHE

Documentation Coverage

This is an auto-generated dedicated concept page to ensure 100% documentation coverage across the ecosystem.

Topological Reasoning Detection (CONCEPT:AHE-3.4)

Overview

Maps user queries to MathematicalFoundationNode or quantitative financial concepts to trigger reasoning models.

Implementation Details

  • Source Code: agent_utilities/graph/_router_impl.py
  • Pillar: AHE

Documentation Coverage

This is an auto-generated dedicated concept page to ensure 100% documentation coverage across the ecosystem.