and now with llm

Dockerfile

@@ -3,14 +3,14 @@ FROM rocker/geospatial:latest
 WORKDIR /code
 
 RUN install2.r --error \
+    bslib \
     shiny \
-    dplyr \
-    ggplot2 \
-    readr \
-    ggExtra \
-    duckdbfs
+    shinychat \
+    tidyverse \
+    duckdbfs \
+    markdown
 
-RUN installGithub.r cboettig/mapgl
+RUN installGithub.r cboettig/mapgl tidyverse/ellmer
 
 COPY . .
 

app.R

@@ -1,32 +1,101 @@
 library(shiny)
 library(bslib)
+library(markdown)
+library(shinychat)
 library(mapgl)
+library(tidyverse)
+library(duckdbfs)
 
 pmtiles <- "https://data.source.coop/cboettig/us-boundaries/mappinginequality.pmtiles"
 
 # Define the UI
 ui <- page_sidebar(
 
-  titlePanel("Demo App"),
   sidebar = sidebar(
+
+    textAreaInput("chat", 
+      "Ask me a question!", 
+      value = "Which state has the highest average social vulnerability?",
+      width = "100%",
+      height = 100
+    ),
+
+    verbatimTextOutput("sql_code"),
+    textOutput("explanation"),
+
     input_switch("redlines", "Redlined Areas"),
     input_switch("svi", "Social Vulnerability", value = TRUE),
     input_switch("richness", "Biodiversity Richness"),
     input_switch("rsr", "Biodiversity Range Size Rarity"),
+  width = 300,
+  ),
+  titlePanel("Demo App"),
 
-    ),
-    # Create a main panel
-  card(full_screen = TRUE,
-      maplibreOutput("map")
+  layout_columns(
+    card(maplibreOutput("map")),
+    card(includeMarkdown("## Plot"),
+         plotOutput("chart1"),
+         plotOutput("chart2"),
+         ),
 
-    )
-  )
+    col_widths = c(8,4)
+  ),
+
+  tableOutput("table")
+
+)
+
+svi <- "https://data.source.coop/cboettig/social-vulnerability/svi2020_us_tract.parquet" |>
+      open_dataset(tblname = "svi")
+
+con <- duckdbfs::cached_connection()
+schema <- DBI::dbGetQuery(con, "PRAGMA table_info(svi)")
+#schema <- svi |> head() |> collect() |> str()
+
+system_prompt = glue::glue('
+You are a helpful agent who always replies strictly in JSON-formatted text.
+Your task is to translate the users question into a SQL query that will be run
+against the "svi" table in a duckdb database. The duckdb database has a
+spatial extension which understands PostGIS operations as well.
+
+The table schema is <schema>
+
+The column called "RPL_THEMES" corresponds to the overall "Social vulnerability index" number. 
+
+Format your answer as follows:
+
+{
+"query": "your raw SQL response goes here",
+"explanation": "your explanation of the query"
+}
+', .open = "<", .close = ">")
 
 # Define the server
 server <- function(input, output, session) {
-  output$map <- renderMaplibre({
 
-    m <- maplibre(center=c(-92.9, 41.3), zoom=4)
+  chat <- ellmer::chat_vllm(
+    base_url = "https://llm.nrp-nautilus.io/",
+    model = "llama3",
+    api_key = Sys.getenv("NRP_API_KEY"),
+    system_prompt = system_prompt
+  )
+
+  observeEvent(input$chat, {
+    stream <- chat$chat(input$chat)
+
+    chat_append("chat", stream)
+    response <- jsonlite::fromJSON(stream)
+
+    output$sql_code <- renderText({response$query})
+    output$explanation <- renderText({response$explanation})
+
+    df <- DBI::dbGetQuery(con, response$query)
+    output$table <- renderTable(df, striped = TRUE)
+
+  })
+
+  output$map <- renderMaplibre({
+    m <- maplibre(center=c(-92.9, 41.3), zoom=3)
 
     if (input$redlines) {
       m <- m |>
@@ -44,7 +113,7 @@ server <- function(input, output, session) {
                           tiles = "https://data.source.coop/cboettig/mobi/tiles/red/species-richness-all/{z}/{x}/{y}.png",
                           maxzoom = 11
                           ) |>
-        add_raster_layer(id = "richness-layer", 
+        add_raster_layer(id = "richness-layer",
                          source = "richness")
 
     }
@@ -72,15 +141,22 @@ server <- function(input, output, session) {
                                   stops = c("lightblue", "darkblue"),
                                   na_color = "lightgrey")
         )
-    }
-
-
 
+
+    }
+  m})
 
 
-    m
-  })
+    chart1 <-
+      svi |>
+      filter(RPL_THEMES > 0) |>
+      group_by(COUNTY) |>
+      summarise(mean_svi = mean(RPL_THEMES)) |>
+      collect() |>
+      ggplot(aes(mean_svi)) + geom_density()
 
+  output$chart1 <- renderPlot(chart1)
+  output$chart2 <- renderPlot(chart1)
 
 }