Module backend 

Backend abstraction. See plan/V2_AGENT_SPEC.md §5.

Each adapter implements Backend, exposing a uniform API over the supported inference engines (vLLM, llama.cpp, LM Studio, Ollama) and remote BYOK passthroughs (OpenRouter, Venice). The agent’s discovery and (in a later task) job-executor layers consume backends through this trait.

Re-exports

pub use llamacpp::LlamaCppBackend;

pub use lmstudio::LmStudioBackend;

pub use ollama::OllamaBackend;

pub use openrouter::OpenRouterBackend;

pub use venice::VeniceBackend;

pub use vllm::VllmBackend;

Modules

llamacpp

llama.cpp server adapter. OpenAI-compatible.

lmstudio

LM Studio adapter. OpenAI-compatible.

ollama

Ollama adapter.

openrouter

OpenRouter BYOK passthrough.

venice

Venice.ai BYOK passthrough.

vllm

vLLM backend adapter. See research/inference-engines.md for details.

Structs

BackendHealth

Health-check result used by discovery.

BackendModel

A model exposed by a backend.

Job

A unit of work received from the coordinator.

JobResult

Result returned to the coordinator when a job completes successfully.

Enums

BackendError

Errors that can be returned at the backend boundary.

WireFormat

Wire format expected by the upstream client. The agent does not transcode between these — it picks a backend that natively speaks the required shape.

Traits

Backend

JobSink

Sink used by adapters to push streaming chunks back to the coordinator.

Type Aliases

BackendResult