Struct Backend 

pub struct Backend {
    pub kind: Kind,
    pub base_url: String,
    pub model: String,
    /* private fields */
}

A multi-provider LLM backend.

Supports Ollama (via OpenAI-compatible API), OpenAI, and Anthropic. All three providers use the same internal Message/ToolCall format — Anthropic’s content-block schema is translated at the wire boundary.

Example

use agnt_net::Backend;

let ollama = Backend::ollama("gemma4:e4b");
let openai = Backend::openai("gpt-4o-mini", "sk-...");
let anthropic = Backend::anthropic("claude-sonnet-4-6", "sk-ant-...");

Fields

kind: Kind

Which provider schema to use on the wire.

base_url: String

Base URL for the provider’s API.

model: String

Model identifier passed in every request.

Implementations

impl Backend

pub fn ollama(model: &str) -> Backend

Create a backend pointing at a local Ollama server.

Uses http://localhost:11434/v1 by default (the OpenAI-compatible endpoint).

pub fn openai(model: &str, api_key: &str) -> Backend

Create a backend for OpenAI’s API.

pub fn anthropic(model: &str, api_key: &str) -> Backend

Create a backend for Anthropic’s native API.

Message format is automatically translated to Anthropic’s content-block schema at the wire boundary — you still work with the OpenAI-style Message type internally.

pub fn with_timeouts( self, connect: Duration, read: Duration, ) -> Result<Backend, String>

Override the HTTP timeouts for this backend instance.

Builds a fresh ureq Agent with the supplied connect/read timeouts and attaches it to this Backend. Subsequent requests made via this instance will use the custom Agent instead of the process-wide shared one.

Returns an error if TLS initialization fails.

pub fn chat( &self, messages: &[Message], tools: &Value, sink: Option<&mut dyn FnMut(&str)>, ) -> Result<Message, String>

Trait Implementations

impl Clone for Backend

fn clone(&self) -> Backend

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Backend

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl LlmBackend for Backend

fn model(&self) -> &str

The model identifier this backend is configured for (e.g. gemma4:e4b).

fn chat( &self, messages: &[Message], tools: &Value, on_token: Option<&mut dyn FnMut(&str)>, ) -> Result<Message, BackendError>

Run one inference turn with the given conversation and tool schemas.