RFC-0006 LLM Provider

RFC-0006 — LLM Provider Interface

Defines the provider abstraction.

Every model backend must implement:

class LLMProvider:

def generate(self, prompt: str) -> str:
    ...

Optional capabilities (detected at runtime):

• stream(prompt, on_token) / astream(...) — incremental output
• agenerate(prompt) — async generation

Bundled adapters (xyberos/llm/adapters.py): OpenAILLM, AnthropicLLM, and GeminiLLM (lazy SDK imports), plus OllamaLLM and OpenAICompatibleLLM (stdlib HTTP).

The Brain depends only on this interface.