sgns::neoswarm::elm::DomainELM¶
Domain-specific ELM (Math, Code, Science) with optional dedicated .mnn model file. More...
#include <domain_elm.hpp>
Inherits from sgns::neoswarm::elm::IELM
Public Functions¶
| Name | |
|---|---|
| DomainELM(ELMRole role, std::shared_ptr< core::InferenceEngine > sharedEngine =nullptr) Construct a DomainELM for the given domain role. |
|
| virtual std::string | GetName() const override |
| virtual ELMRole | GetRole() const override |
| virtual bool | IsLoaded() const override |
| virtual outcome::result< void > | Load(const std::string & model_path) override Load the ELM model from disk (if a dedicated model path is configured). |
| virtual outcome::result< std::string > | Process(const std::string & input, const ELMContext & context) override Process input through this ELM and return refined output. |
Additional inherited members¶
Public Functions inherited from sgns::neoswarm::elm::IELM
| Name | |
|---|---|
| virtual | ~IELM() =default |
| virtual float | GetConfidence() const Confidence in the last Process() call. |
Protected Attributes inherited from sgns::neoswarm::elm::IELM
| Name | |
|---|---|
| float | m_lastConfidence |
Detailed Description¶
Domain-specific ELM (Math, Code, Science) with optional dedicated .mnn model file.
Per D-01/D-02: by default uses the shared MNN backbone; if Load() is called with a non-empty model_path, creates a dedicated MNNInferenceEngine instance and uses that instead. Fail-close per D-04.
Public Functions Documentation¶
function DomainELM¶
Construct a DomainELM for the given domain role.
Parameters:
- role ELMRole::Math, Code, or Science.
- sharedEngine Shared inference engine (MNN backbone). May be nullptr for stub/testing; Process will fail-close.
function GetName¶
Return: Human-readable name of this ELM.
Reimplements: sgns::neoswarm::elm::IELM::GetName
function GetRole¶
Return: The ELM role this instance fulfills.
Reimplements: sgns::neoswarm::elm::IELM::GetRole
function IsLoaded¶
Return: True if the ELM model has been loaded.
Reimplements: sgns::neoswarm::elm::IELM::IsLoaded
function Load¶
Load the ELM model from disk (if a dedicated model path is configured).
Parameters:
- model_path Path to the model file (.mnn or similar).
Return: outcome::success or ModelLoadFailed.
Reimplements: sgns::neoswarm::elm::IELM::Load
function Process¶
virtual outcome::result< std::string > Process(
const std::string & input,
const ELMContext & context
) override
Process input through this ELM and return refined output.
Parameters:
- input Text to process (typically output of previous chain step).
- context ELM execution context (original task, prior outputs, confidences).
Return: Refined text or InferenceFailed.
Reimplements: sgns::neoswarm::elm::IELM::Process
Updated on 2026-10-09 at 03:59:42 +0000