2026-03-21.22-55-17.mp4
2026-03-30.01-08-27.mp4
pip install -r requirements.txtIf you want the true inline chat experience in VS Code, follow:
VS_CODE_MCP_SETUP.md
python server.pyThis server now builds on panel-live-server, so compatible MCP clients can render
create_plot and update_plot inline in the chat pane instead of only returning a URL.
For reliability, the structured tools use a lightweight HoloViews rendering path for single plots,
while interactive widget apps can be rendered through show or interactive_sine_wave.
python test_local.pyThis simulates MCP tool calls locally, prints the returned payload, and opens the fallback
browser URLs. It also includes an interactive HoloViews sine-wave app with sliders. Inline
rendering only happens inside a compatible MCP client UI.
Keep the terminal open while viewing the demo. The /feed endpoint is not the visualization;
open the returned /view?id=... URL instead.
Example create_plot input:
{
"data": [
{"week": 1, "sales": 120},
{"week": 2, "sales": 135},
{"week": 3, "sales": 128},
{"week": 4, "sales": 150},
{"week": 5, "sales": 162}
],
"plot_type": "scatter",
"x": "week",
"y": "sales"
}Example update_plot input:
{
"plot_id": "<returned plot_id>",
"data": [
{"week": 1, "sales": 120},
{"week": 2, "sales": 135},
{"week": 3, "sales": 128},
{"week": 4, "sales": 150},
{"week": 5, "sales": 162},
{"week": 6, "sales": 174},
{"week": 7, "sales": 181}
],
"plot_type": "line"
}When you run python test_local.py, it uses this exact scenario and prints the MCP payloads
returned by both tools.
When the user wants a plot rendered inline in chat with widget controls, call:
interactive_sine_wave
This tool returns a Panel app backed by HoloViews and includes sliders for:
amplitudefrequencyphase
For custom widget-driven apps, use:
show
create_plotupdate_plotlist_plot_types
showvalidatelist_packages
scatterlinebar