Function
Coordinate workflow
Run related tasks with explicit ordering, state, and operational handling.
Studio / Area
Composition & Runtime
Assemble components and operate the flow of work between them.
How are components assembled, invoked, coordinated, and operated?
In scope
- Serving, orchestration, routing, interfaces, queues, schedules, and deployment
Boundary
- The identity of a model release; authority to approve an action
Start with the outcome
Functions
Function
API contract description
Describe HTTP operations, parameters, and data shapes in a shared interface format.
Function
Rule evaluation
Match input facts against declared rule conditions and execute applicable outcomes.
Function
Tool invocation through a shared protocol
Expose named tools through a defined client-server interface.
Function
Workflow orchestration
Coordinate and observe tasks that make up a repeatable process.
Function
Drools
Rule engine for evaluating declared business rules against facts.
Function
Prefect
Workflow orchestration software for building and operating task flows.
Function
Model Context Protocol (2025-11-25)
Versioned protocol defining how servers expose prompts, resources and tools.
Function
OpenAPI Specification 3.2.1
Specification for describing HTTP APIs in a language-independent form.
Function
Describe a service interface
Make an HTTP service’s operations and inputs understandable to callers.
Function
Apply explicit rules
Evaluate stated conditions against facts to produce a defined decision or action.
Function
Invoke an external tool
Call a declared tool through an application-controlled interface and handle its result.
Function
Serve model inference
Expose a model through a request endpoint for application use.
Function
Model serving
Run a deployed model behind an API that applications can call.
Function
vLLM
Inference engine with an HTTP server for supported model capabilities.
Function
Adapt a model
Adjust a pretrained model for a new task or domain using additional training.
Function
Local model serving
Run a model on a user-controlled machine and expose a local inference endpoint.
Function
Parameter-efficient fine-tuning
Train a limited set of parameters or adapters while keeping most base weights fixed.
Function
Low-rank adaptation (LoRA)
Train low-rank adapter matrices for selected model weights.
Function
Ollama local runtime
Locally installed Ollama runtime for downloading and serving supported models.
Function
Hugging Face PEFT
Python library for parameter-efficient model adaptation, including LoRA.
Representative paths
Methods and options
approach
Workflow orchestration
Coordinate and observe tasks that make up a repeatable process.
implementation
Prefect
Workflow orchestration software for building and operating task flows.
approach
API contract description
Describe HTTP operations, parameters, and data shapes in a shared interface format.
standard
OpenAPI Specification 3.2.1
Specification for describing HTTP APIs in a language-independent form.
approach
Rule evaluation
Match input facts against declared rule conditions and execute applicable outcomes.
implementation
Drools
Rule engine for evaluating declared business rules against facts.
approach
Tool invocation through a shared protocol
Expose named tools through a defined client-server interface.
standard
Model Context Protocol (2025-11-25)
Versioned protocol defining how servers expose prompts, resources and tools.
approach
Model serving
Run a deployed model behind an API that applications can call.
implementation
vLLM
Inference engine with an HTTP server for supported model capabilities.
approach
Local model serving
Run a model on a user-controlled machine and expose a local inference endpoint.
approach
Parameter-efficient fine-tuning
Train a limited set of parameters or adapters while keeping most base weights fixed.
approach
Low-rank adaptation (LoRA)
Train low-rank adapter matrices for selected model weights.
implementation
Ollama local runtime
Locally installed Ollama runtime for downloading and serving supported models.
implementation
Hugging Face PEFT
Python library for parameter-efficient model adaptation, including LoRA.
Across the Studio
Governance connection
Runtime records distinguish what a system can execute from who is allowed to authorize it.
Related records
- In this areaCoordinate workflow
- In this areaAPI contract description
- In this areaRule evaluation
- In this areaTool invocation through a shared protocol
- In this areaWorkflow orchestration
- In this areaDrools
- In this areaPrefect
- In this areaModel Context Protocol (2025-11-25)
- In this areaOpenAPI Specification 3.2.1
- In this areaDescribe a service interface
- In this areaApply explicit rules
- In this areaInvoke an external tool
- In this areaServe model inference
- In this areaModel serving
- In this areavLLM
- In this areaAdapt a model
- In this areaLocal model serving
- In this areaParameter-efficient fine-tuning
- In this areaLow-rank adaptation (LoRA)
- In this areaOllama local runtime
- In this areaHugging Face PEFT
Sources
Last reviewed 2026-09-29