A table can answer a question and still leave you with another job. You have the monthly figures, but want to see the trend. You have a total, but need it to stand out. Then someone asks for a document they can read without opening the conversation.
GenViz can return those forms of an answer from one Python analysis: a chart, a KPI card and a downloadable PDF report. The calculation stays beside the results, so you can inspect how each number was produced. That’s conversational analytics: ask in context and examine the answer where it appears.
See the shape of the data
Ask for revenue by month and a line chart can show the movement. Compare regions and a bar chart can make the differences easier to see. A scatter plot can help examine how two measurements vary together. Each chart appears in the conversation with the rows behind it available for inspection.
The assistant writes and runs the Python analysis that prepares the result. Its Code tab shows the calculation, while the conversation shows the returned values and visuals. For example, a summary with revenue of 100 in January, 200 in February and 300 in March can produce three plotted points and a total of 600 from the same data.
Bring a key number forward
A KPI card gives one scalar its own space, with a title and number formatting such as currency or percent. A value of 600 formatted as USD appears as $600.00. A fraction of 0.125 formatted as a percent appears as 12.5%. These are separate KPI results; the underlying Python calculation determines what each one means.
KPI cards are distinct from report callout cards. A KPI presents a formatted number. A callout carries a heading and explanatory text. Ordinary named values and tables can appear alongside chart and KPI results when the answer also needs its underlying figures.
Take a PDF snapshot
Ask for a PDF when the answer needs to travel. A report can combine a short explanation, key figures, charts and small supporting tables. Its result shows a title, page count and download action in the conversation. On desktop, Download opens the native save dialog; cancelling leaves the report ready to download again.
The PDF captures that analysis. If the source conversation later gets a new result, an already saved copy still contains the figures from when it was created. Run a new analysis to make an updated report. You can also save the conversation result to yourself and reopen that saved snapshot; its report remains available as its own copy.
The assistant’s code can return typed results like these:
chart = genviz.Chart(
summary, x="month", series=[{"key": "revenue", "type": "line"}],
title="Monthly revenue",
)
total = genviz.KPI(600, format="currency", currency="USD", title="Revenue")
rate = genviz.KPI(0.125, format="percent", decimals=1, title="Conversion rate")
genviz.result(chart)
genviz.result(total)
genviz.result(rate)
genviz.result(genviz.Report(
"Revenue review",
sections=[total, chart],
))The example shows the shapes of the returned data. The assistant prepares the summary from the question and available inputs; you can inspect its code when you want to follow the calculation.
Keep results together
Charts and reports have size limits, so an oversized result needs a smaller sample or an aggregation. A report can include a small table to support its summary; a full set of records still belongs in a downloadable extract.
If one report cannot be produced, GenViz shows that report as unavailable while successful charts, KPI cards, values and tables remain in the conversation. You can refine the request with the useful work still in view. This keeps conversational analytics connected to the calculation and its results.