Intelligently optimize and compress context for LLM prompts to greatly increase effective context window
