Orchestration · Software component

Function-Calling Controller

Software componentOrchestrationOrchestration & Toolsarc:FunctionCallingController

An agent controller that presents all registered function specifications to an LLM and lets model-native function calling choose, parameterize, order or parallelize function invocations, adapting to returned results.

Responsibility. Routes each request dynamically to registered functions using LLM function-calling decisions.

Also known as: Kernel orchestrator, Orchestrator LLM, Dynamic plugin router, Function-calling planner, Auto function invocation

Variant of Agent Controller abstract

When to choose. Choose when an agent must integrate an extensive, growing catalog of capabilities (tens of functions across many domains) whose composition cannot be pre-programmed and new capabilities must be added without changing orchestration code. Avoid for focused agents with fewer than five capabilities, or for workflows needing iterative refinement loops, conditional branching on intermediate results, or strictly deterministic execution (prefer explicit state-graph orchestration).

Direct neighbourhood (hover for relationship types)

Relationships

is configured by structural

Tool Schema Ch2.5 Ref7.13

invokes dependency

reads dependency

Tool Registry Ch2.5

emits telemetry to dynamic

is constrained by control

Function Choice Policy Ch2.5

Design guidance

SHOULD incorporate newly registered functions into routing without changes to orchestration code.
MUST NOT be relied on for deterministic execution sequences; routing is non-deterministic even at temperature 0, so compliance-critical steps need constrained function-choice behaviour.
SHOULD serve as the single mediation point where monitoring, token accounting and audit of executed capabilities are injected uniformly.

Quantitative guidance

As stated by the sources; verify before use.

Plugin routing is described as scaling from about 5 to 50 capabilities (10 to 50 functions) without proportional orchestration complexity (Ch2.5).
Routing identical queries can differ across executions even at temperature=0 (Ch2.5).

Classification

Patterns: LLM-driven dynamic routingAutomatic function callingPlugin architectureSemantic-native function composition
Technologies: Semantic KernelAzure OpenAI Service
Quality attributes: Maintainability (ISO/IEC 25010)Flexibility (ISO/IEC 25010)
Risks mitigated: Orchestration code changes for every new capabilityUnmaintainable hand-coded routing as capability count grows

Sources

Ch2.5: T. Nguyen, "Semantic Kernel," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 2.5. ISBN: 9798244538229.
Ref7.13: NVIDIA, "Llama Nemotron," NVIDIA NeMo Framework User Guide, v25.09. Accessed: Sep. 27, 2026. [Online]. Available: https://docs.nvidia.com/nemo-framework/user-guide/25.09/llms/llama_nemotron.html