AI Overview

RFlect has optional AI integration on two surfaces:

1. In-GUI — AI Chat Assistant + AI-powered DOCX reports
2. MCP — agent clients (Claude Code, Cline, …) drive RFlect programmatically via 34 tools

Both are powered by the same unified provider abstraction in plot_antenna/llm_provider.py, which supports OpenAI, Anthropic, and Ollama.

Always optional

AI features are completely optional. Without a configured provider:

• The AI Chat menu item is hidden
• Generate Report with AI falls back to template-only mode
• Core RFlect functionality (plotting, calculation, MCP tool calling) continues to work normally

What AI is good at here

• Executive summaries in reports — "this antenna's peak gain is 5.2 dBi at 2450 MHz; the pattern is mildly directional with HPBW of 78° in E-plane"
• Pattern interpretation — comparing HPOL vs VPOL, calling out polarization mismatches
• Conversational queries over loaded measurements — "what's the worst-case gain in the 2.4 GHz band?"
• Vision over rendered plots — given a 3D pattern PNG, describe what the operator should notice

What AI is not doing

• The math. All gain, TRP, polarization, and bandwidth calculations are pure-Python in plot_antenna/calculations.py and plot_antenna/ai_analysis.py. The AI layer only consumes the numbers.
• Hardware control. RFlect doesn't drive instruments.
• Replacing engineering judgment. The disclaimer in AI Status applies.

Where to next

• Status — current state, what works, what's pending
• Providers — configure OpenAI / Anthropic / Ollama
• Chat — using the in-GUI AI Chat Assistant
• Reports — AI-augmented DOCX reports