Backend metadata for introspection. More...

#include <entropic/types/backend_capability.h>

Public Attributes
std::string	name
	Backend identifier (e.g. "llama.cpp", "axcl")

std::string	version
	Backend version string.

std::string	compute_device
	"cuda", "vulkan", "cpu", "npu"

std::string	model_format
	"gguf", "axmodel", "onnx", etc.

std::string	architecture
	Architecture family of the loaded model.

int	max_context_length = 0
	Maximum context length.

size_t	vram_bytes = 0
	VRAM consumed by loaded model (bytes). 0 if COLD.

size_t	ram_bytes = 0
	RAM consumed by loaded model (bytes). 0 if COLD.

size_t	parameter_count = 0
	Number of parameters (from model metadata).

std::string	quantization
	Quantization type (e.g. "IQ3_XXS", "Q8_0", "fp16").

Detailed Description

Backend metadata for introspection.

Returned by InferenceBackend::info(). The engine uses this for logging, diagnostics (v1.9.12), and routing decisions.

Version: 1.9.13

Definition at line 58 of file backend_capability.h.

Member Data Documentation

◆ architecture

std::string entropic::BackendInfo::architecture

Architecture family of the loaded model.

"transformer", "gdn", "mamba", "rwkv", "hybrid", "unknown" Populated after load(). Empty when COLD.

Version: 1.9.13

Definition at line 68 of file backend_capability.h.

◆ compute_device

std::string entropic::BackendInfo::compute_device

"cuda", "vulkan", "cpu", "npu"

Definition at line 61 of file backend_capability.h.

◆ max_context_length

int entropic::BackendInfo::max_context_length = 0

Maximum context length.

For transformers: fixed window from model metadata. For recurrent: -1 (theoretically unlimited, practically memory-bound).

Version: 1.9.13

Definition at line 74 of file backend_capability.h.

◆ model_format

std::string entropic::BackendInfo::model_format

"gguf", "axmodel", "onnx", etc.

Definition at line 62 of file backend_capability.h.

◆ name

std::string entropic::BackendInfo::name

Backend identifier (e.g. "llama.cpp", "axcl")

Definition at line 59 of file backend_capability.h.

◆ parameter_count

size_t entropic::BackendInfo::parameter_count = 0

Number of parameters (from model metadata).

Definition at line 78 of file backend_capability.h.

◆ quantization

std::string entropic::BackendInfo::quantization

Quantization type (e.g. "IQ3_XXS", "Q8_0", "fp16").

Definition at line 79 of file backend_capability.h.

◆ ram_bytes

size_t entropic::BackendInfo::ram_bytes = 0

RAM consumed by loaded model (bytes). 0 if COLD.

Definition at line 77 of file backend_capability.h.

◆ version

std::string entropic::BackendInfo::version

Backend version string.

Definition at line 60 of file backend_capability.h.

◆ vram_bytes

size_t entropic::BackendInfo::vram_bytes = 0

VRAM consumed by loaded model (bytes). 0 if COLD.

Definition at line 76 of file backend_capability.h.

The documentation for this struct was generated from the following file:

entropic/types/backend_capability.h

Public Attributes