Digikam::SearchLlamaWorker Class Reference

Worker object living on the worker thread. More...

Inheritance diagram for Digikam::SearchLlamaWorker:

Public Slots
void	slotDoInference (const QString &prompt, int maxTokens, float temperature)
void	slotDoLoad (const QString &modelPath)
void	slotDoUnload ()

Signals
void	signalCancelled ()
void	signalError (const QString &message)
void	signalLoaded (bool success)
void	signalLoadProgress (int percent)
void	signalOutputReady (const QString &output)
void	signalProgress (int step)

Public Member Functions
	SearchLlamaWorker (QObject *const parent=nullptr)
bool	abortRequested () const
	Thread-safe check used by the llama.cpp abort callback to stop an in-progress decode when cancellation has been requested.
void	reportLoadProgress (float progress)
	Relay model-load progress (0.0-1.0) from the llama.cpp progress callback to the GUI as a percentage.
void	requestCancel ()

Detailed Description

Owns the llama.cpp model and context. All llama_* calls happen here, never on the GUI thread.