AI Chat Assistant¶

In-GUI conversational analysis of loaded measurements.

Opening it¶

Tools → AI Chat (only visible when a provider is configured).

The chat window has: - Conversation pane - Input box - Quick-action buttons: Gain Stats, Pattern, Polarization, All Freqs

Function calling¶

The chat assistant has five callable functions wired to the AI:

Function	Returns
`get_gain_statistics()`	Min / max / spherical avg gain
`analyze_pattern()`	HPBW, F/B ratio, nulls, sidelobes
`compare_polarizations()`	AR, tilt, XPD, sense
`generate_2d_plot()`	2D pattern description (textual)
`generate_3d_plot()`	3D pattern description (textual)

The model picks tools based on your question and chains them as needed.

Multi-turn context¶

Conversation history is preserved across messages. The assistant remembers which measurement you loaded and which frequency you've been discussing.

Example prompts¶

"What's the peak gain at 2.45 GHz?"
"Compare HPOL vs VPOL at this frequency."
"Is the pattern more directional or more omnidirectional?"
"Summarize the antenna's performance across all loaded frequencies."

Vision¶

If your provider supports vision (GPT-4o+, Claude, llava/llama3.2-vision), you can ask the model to interpret a rendered plot. The chat window's "attach plot" button base64-encodes the current matplotlib figure and includes it in the message.

Limits¶

Older Ollama models may have limited tool-calling support
Pattern analysis is currently limited to three pattern types
No sidelobe-level detection yet

See AI Status for the full list.

Programmatic access¶

The same analyses are available via MCP without the GUI:

get_gain_statistics(2450.0)
analyze_pattern(2450.0, "total")
compare_polarizations(2450.0)

See MCP Tools Reference.