Crate aonyx_llm 

aonyx-llm

Multi-provider LLM router. One LlmProvider trait, several implementations, one configurable fallback chain.

Providers

• anthropic — native Anthropic Messages API (streaming SSE).
• openai_compat — shared OpenAI-compatible backend.
  • openai — public OpenAI API (https://api.openai.com).
  • openrouter — OpenRouter aggregator, with optional attribution headers.
  • lm_studio — local OpenAI-compatible LM Studio server.
• ollama — local Ollama (/api/chat), JSON-lines streaming.
• claude_code — wraps the installed claude binary; no API key required (auth delegated to Claude Code subscription / its env vars).
• nous_portal — Nous Portal endpoint (deferred).

Router

Router holds an ordered list of providers and forwards each request to the first one whose stream opens successfully.

Re-exports

pub use claude_code::ClaudeCodeProvider;

pub use claude_code::CLAUDE_DEFAULT_BIN;

pub use ollama::OllamaProvider;

pub use ollama::OLLAMA_DEFAULT_BASE_URL;

pub use openai_compat::OpenAiCompatProvider;

pub use router::Router;

Modules

anthropic

Anthropic Claude provider — streaming over the Messages API.

claude_code

Claude Code provider — uses an installed claude binary as the backend.

lm_studio

LM Studio provider — OpenAI-compatible with a user-configurable base URL.

nous_portal

Nous Portal provider.

ollama

Ollama provider — JSON-lines streaming from POST /api/chat.

openai

OpenAI Chat Completions provider.

openai_compat

OpenAI Chat Completions compatible providers.

openrouter

OpenRouter provider.

retry

Shared HTTP retry/backoff for LLM providers (Phase RR).

router

Fallback chain router.