ML.Agentic & ML.Reasoning Primitives
MACHINE LEARNING SUITE RELEASED V1.0.0 LOCAL AIR-GAPPED

ML.Agentic & ML.Reasoning Primitives

21 zero-dependency building blocks, MCP, and A2A protocols for multi-agent systems and 5 graph-grounded reasoning engines.

#Agentic#Reasoning#MCTS#BDI#Guardrails#Releases

ML.Agentic & ML.Reasoning Primitives

ML.Agentic and ML.Reasoning are open-core libraries providing standalone, zero-dependency architectural building blocks for multi-agent coordination, goal planning, and graph-grounded verification.

Package Installation

# Add ML.Agentic to your .NET 10 project
dotnet add package HFLabs.ML.Agentic --version 1.0.0

22 ML.Agentic Primitives

┌─────────────────────────────────────────────────────────────────────────┐
│                        22 ML.Agentic Primitives                         │
├───────────────────────┬─────────────────────────┬───────────────────────┤
│ Planning & Search     │ Reasoning & Verification│ Coordination & Routing│
│ • Hierarchical Planner│ • Reflection Loop       │ • Blackboard System   │
│ • MCTS Exploration    │ • Chain-of-Verification │ • Semantic Router     │
│ • Priority Scheduler  │ • Multi-Persona Debate  │ • Topology Manager    │
│ • Goal-Progress Mon.  │ • Program-Aided Reason. │ • Fallback Chain      │
│                       │ • Judge Evaluator       │ • Tool Registry       │
│                       │                         │ • Agent Cards Manifest│
├───────────────────────┴─────────────────────────┴───────────────────────┤
│ Safety, Memory & Adaptation                                             │
│ • Guardrail Pipeline  • Human-in-the-Loop Gate   • Resource Budget Ledger│
│ • Memory Bank (Epis.) • Context Compactor        • Evolutionary Mutator │
└─────────────────────────────────────────────────────────────────────────┘

5 ML.Reasoning Graph Algorithms

These algorithms execute against any graph implementation conforming to IKnowledgeGraph:

  1. Multi-Hop Causal Path Reasoning: Discovers directed causal inference chains with confidence decay calculations.
  2. Grounding Verifier: Cross-checks agent claims against verified subgraphs, identifying unsupported assertions.
  3. TAPE Rollout Planning: Predictive state rollout exploring action trajectories before committing execution steps.
  4. Belief-Desire-Intention (BDI) Engine: Maintains formal agent world state, intention sets, and execution plans.
  5. Cognitive-Bias Detection & Debiasing: Detects confirmation bias, anchoring heuristics, and availability bias in agent chains.

Hardware Acceleration

ML.Agentic includes native hooks for optional hardware compute offload via LevelZero.NET. When Intel GPU hardware is detected, semantic intent routing and property graph grounding queries execute via precompiled Level Zero SPIR-V kernels with sub-millisecond dispatch times.

Resources & Links