Add 5 MCP tools (PatternImportMixin) and 1 prompt for importing external
antenna pattern data. Pure-Python parsers with IDW interpolation on the
sphere, single-cut-to-3D synthesis for EMCAR/2-col CSV, and Touchstone S1P
bridge to the analytical pattern engine. Web UI gets a "Load File" button
with multipart upload endpoint and WebSocket broadcast. 78 tools, 14 prompts.
Add analytical radiation pattern models for 5 antenna types (dipole, monopole,
EFHW, loop, patch) driven by S11 impedance measurements. Pure Python math with
closed-form far-field equations — no numpy or simulation dependencies.
New MCP tools:
- radiation_pattern: scan S11 → find resonance → compute 3D pattern
- radiation_pattern_from_data: compute from known impedance (no hardware)
- radiation_pattern_multi: patterns across a frequency band for animation
Web UI (opt-in via MCNANOVNA_WEB_PORT env var):
- Three.js gain-mapped sphere with OrbitControls
- Surface/wireframe/plane cut display modes with teal→amber color ramp
- Smith chart overlay, dBi reference rings, E/H plane cross-sections
- Real-time WebSocket push on new pattern computation
- FastAPI backend shares process with MCP server, zero new core deps
Frontend: Vite + TypeScript + Three.js, built assets committed to webui/static/.
Optional dependencies: fastapi + uvicorn via pip install mcnanovna[webui].
Convert scan, analyze, capture, and cal to async with MCP progress
reporting via FastMCP Context. Blocking serial I/O wrapped with
asyncio.to_thread() so the event loop stays free to deliver progress
notifications during hardware sweeps.
Add connection keepalive: _ensure_connected validates stale connections
with a sync probe after 30s idle, and retries on cold-start failures
(fixes flaky first-connect after MCP server restart).
- Add Pillow dependency for server-side RGB565→PNG conversion
- capture() now returns fastmcp Image (PNG) by default
- Pass raw=True to get the original RGB565 base64 dict
- Eliminates 204K token overflow on default capture calls
