Architecture Overview

This document explains the separation of concerns in YAMLLM.

• LLM (yamllm.core.llm):
  • Loads config, owns request settings, memory, and tool orchestration
  • Selects a provider via yamllm.core.providers.ProviderFactory
  • Exposes streaming via set_stream_callback so callers control rendering
  • Delegates tool definitions and execution to ToolManager
• Providers (yamllm/core/providers/*):
  • Implement a single BaseProvider interface (completion, streaming, embeddings, tool formatting)
  • Contain no UI or printing; return SDK-native objects or iterators
• Tooling (yamllm/tools/*):
  • Tool base class provides schema exposure via get_signature()
  • ToolManager registers tools, exposes provider-friendly definitions, and executes with timeouts
• UI (yamllm/ui/rich_renderer.py):
  • Optional helper that attaches a streaming renderer to an LLM instance
  • Keeps all rendering out of core logic