:construction: :construction:

app.R

@@ -3,16 +3,18 @@ library(bslib)
 library(htmltools)
 library(fontawesome)
 library(bsicons)
-library(gt)
-library(colourpicker)
 library(glue)
-
-library(ggplot2)
-library(readr)
+library(sf)
+library(duckdb.agent)
+library(duckdbfs)
 library(dplyr)
+library(ellmer)
 library(mapgl)
-library(duckdbfs)
-duckdbfs::load_spatial()
+library(digest)
+library(stringr)
+library(shinybusy)
+
+duckdbfs::close_connection()
 
 css <-
   HTML(paste0("<link rel='stylesheet' type='text/css' ",
@@ -26,63 +28,51 @@ ui <- page_sidebar(
   fillable = FALSE, # do not squeeze to vertical screen space
   tags$head(css),
   titlePanel("Demo App"),
+  shinybusy::add_busy_spinner(),
+  
+  p("
+  Select a desired area with the draw tools on the map, then hit 'Set Area of Interest' to select.
+  Then, enter your query in the text box below the map to count occurrences of your specified taxonomic group.
+  Use the airplane button to sned your query.  The computation may take a few minutes depending on the size and scale of
+  the search.
+"),
+
+p("
+Scroll to zoom, ctrl+click to pitch and rotate. Hitting the area button with no selection to include the entire map.
+"),
+
+  layout_columns(
+    card(maplibreOutput("map")),
+    div(actionButton("get_features", "Set Area Of Interest", icon = icon("object-group"),
+                   class = "btn-primary align-bottom")),
+    col_widths = c(11,1)
+    ),
 
-  "
-  This is a proof-of-principle for a simple chat-driven interface
-  to dynamically explore geospatial data.
-  ",
 
   card(
     layout_columns(
       textInput("chat",
         label = NULL,
-        "Which four counties in California have the highest average social vulnerability?",
+        "show all bird occurrences at zoom level 6",
         width = "100%"),
+
       div(
       actionButton("user_msg", "", icon = icon("paper-plane"),
                    class = "btn-primary btn-sm align-bottom"),
       class = "align-text-bottom"),
-      col_widths = c(11, 1)),
+      col_widths = c(11, 1),
       fill = FALSE
+    ),
   ),
   
   textOutput("agent"),
 
-
-  layout_columns(
-    card(maplibreOutput("map")),
-    card(includeMarkdown("## Plot"),
-         plotOutput("chart1"),
-         plotOutput("chart2"),
-         ),
-    col_widths = c(8, 4),
-    row_heights = c("500px"),
-    max_height = "600px"
-  ),
-
-  gt_output("table"),
-
-  card(fill = TRUE,
-    card_header(fa("robot"),  textOutput("model", inline = TRUE)),
-    accordion(
-      open = FALSE,
-      accordion_panel(
-        title = "show sql",
-        icon = fa("terminal"),
-        verbatimTextOutput("sql_code"),
-      ),
-      accordion_panel(
-        title = "explain",
-        icon = fa("user", prefer_type="solid"),
-        textOutput("explanation"),
-      )
-    ),
-  ),
-  card(
-      card_header("Errata"),
-      shiny::markdown(readr::read_file("footer.md")),
-  ),
   sidebar = sidebar(
+    card(fill = TRUE,
+    card_header("Selected area:"),
+    verbatimTextOutput("feature_output")
+    ),
+
     selectInput(
     "select",
     "Select an LLM:", 
@@ -91,192 +81,164 @@ ui <- page_sidebar(
          "Gorilla (UC Berkeley)" = "gorilla" 
         )
   ),
-
-    input_switch("redlines", "Redlined Areas", value = FALSE),
-    input_switch("svi", "Social Vulnerability", value = TRUE),
-    input_switch("richness", "Biodiversity Richness", value = FALSE),
-    input_switch("rsr", "Biodiversity Range Size Rarity", value = FALSE),
-
-    colourInput("color", "Select a color",
-                value = "blue"),
-    sliderInput("slider", "Show BIR74 values above:",
-                value = 248, min = 248, max = 21588),
-
+    card(fill = TRUE,
+      card_header(fa("robot"),  textOutput("model", inline = TRUE)),
+      accordion(
+        open = TRUE,
+        accordion_panel(
+          HTML("<span, class='text-info'>Show SQL query</span>"),
+          icon = fa("terminal"),
+          verbatimTextOutput("sql_code")
+        ),
+        accordion_panel(
+          title = "Explain query",
+          icon = fa("user", prefer_type = "solid"),
+          textOutput("explanation")
+        )
+      )
+    ),
+   
+   
     card(
       card_header(bs_icon("github"), "Source code:"),
-      a(href = "https://github.com/boettiger-lab/geo-llm-r",
-        "https://github.com/boettiger-lab/geo-llm-r"))
+      a(href = "https://github.com/boettiger-lab/biodiversity-justice",
+        "https://github.com/boettiger-lab/biodiversity-justice"))
   ),
 
   theme = bs_theme(version = "5")
 )
 
 
-repo <- "https://data.source.coop/cboettig/social-vulnerability"
-pmtiles <- glue("{repo}/2022/SVI2022_US_tract.pmtiles")
-parquet <- glue("{repo}/2022/SVI2022_US_tract.parquet")
-con <- duckdbfs::cached_connection()
-svi <- open_dataset(parquet, tblname = "svi") |> filter(RPL_THEMES > 0)
-
-safe_parse <- function(txt) {
-  gsub("[\r\n]", " ", txt) |> gsub("\\s+", " ", x = _)
-}
-
-
-# helper utilities
-# faster/more scalable to pass maplibre the ids to refilter pmtiles,
-# than to pass it the full geospatial/sf object
-filter_column <- function(full_data, filtered_data, id_col = "FIPS") {
-  if (nrow(filtered_data) < 1) return(NULL)
-  values <- full_data |>
-    inner_join(filtered_data, copy = TRUE) |>
-    pull(id_col)
-  # maplibre syntax for the filter of PMTiles  
-  list("in", list("get", id_col), list("literal", values))
-}
-
+duckdb_secrets(Sys.getenv("MINIO_KEY"),
+            Sys.getenv("MINIO_SECRET"),
+            "minio.carlboettiger.info")
 
+gbif <- open_dataset("s3://public-gbif/2024-10-01",  tblname = "gbif")
+bounds <- ""
 
 # Define the server
 server <- function(input, output, session) {
+  output$map <- renderMaplibre({
+    m <- maplibre(center=c(-110, 38), zoom = 3, pitch = 0) |>
+      add_draw_control() |>
+      add_geocoder_control() #|> 
+      # set_projection("globe")
+    
+    m
+  })
+observeEvent(input$get_features, {
 
-  chart1_data <- svi |>
-    group_by(COUNTY) |>
-    summarise(mean_svi = mean(RPL_THEMES)) |>
-    collect()
-
-  chart1 <- chart1_data |>
-    ggplot(aes(mean_svi)) + geom_density(fill="darkred") +
-    ggtitle("County-level vulnerability nation-wide")
-  
-  data <- reactiveValues(df = tibble())
-  output$chart1 <- renderPlot(chart1)
-
-  model <- reactive(input$select)
-  output$model <- renderText(input$select)
-  observe({
-    schema <- read_file("schema.yml")
-    system_prompt <- glue::glue(readr::read_file("system-prompt.md"),
-                                .open = "<", .close = ">")
-    chat <- ellmer::chat_vllm(
-      base_url = "https://llm.nrp-nautilus.io/",
-      model =  model(),
-      api_key = Sys.getenv("NRP_API_KEY"),
-      system_prompt = system_prompt,
-      api_args = list(temperature = 0)
-    )
-
-  observeEvent(input$user_msg, {
-    stream <- chat$chat(input$chat)
+      drawn_features <- get_drawn_features(mapboxgl_proxy("map"))
+      if(nrow(drawn_features) > 0) {
+        bounds <- st_bbox(drawn_features)
 
-    # Parse response
-    response <- jsonlite::fromJSON(safe_parse(stream))
-    #response <- jsonlite::fromJSON(stream)
+        # print(bounds)
+        output$feature_output <- renderPrint({
+            print(bounds)
+        })
 
-    if ("query" %in% names(response)) {
-      output$sql_code <- renderText(stringr::str_wrap(response$query, width = 60))
-      output$explanation <- renderText(response$explanation)
 
-      # Actually execute the SQL query generated:
-      df <- DBI::dbGetQuery(con, response$query)
+        attach(as.list(bounds))
+        gbif_aoi <- gbif |>
+          dplyr::filter(between(decimallatitude, ymin, ymax),
+                      between(decimallongitude, xmin, xmax)) |>
+          as_view("gbif_aoi")
 
-      # don't display shape column in render
-      df <- df |> select(-any_of("Shape"))
-      output$table <- render_gt(df, height = 300)
+      }
 
+  observeEvent(input$user_msg, {
 
-      y_axis <- colnames(df)[!colnames(df) %in% colnames(svi)]
-      chart2 <- df |>
-        rename(social_vulnerability = y_axis) |>
-        ggplot(aes(social_vulnerability)) +
-        geom_density(fill = "darkred")  +
-        xlim(c(0, 1)) +
-        ggtitle("Vulnerability of selected areas")
 
-      output$chart2 <- renderPlot(chart2)
 
-      # We need to somehow trigger this df to update the map.
-      data$df <- df
+    system_prompt = create_prompt(additional_instructions =
+    "Note that the columns h1, h2, h3, through h11 contains a geohash representing a H3 hexagon index.
+    Higher numbers indicate higher zoom resolution (smaller hexes)
+    Always aggregate results to count the number of rows matching
+    the query to the desired hexagon. Always name the count column 'count'.
+    Remember to group by hexagon level to aggregate! 
+    Always rename the chosen hexagon column as 'h3id' in your final answer.
+    Always use table notation like 'gbif.order' to specify column names.
+    Be sure to generate fully valid SQL. Check your SQL for possible errors.
 
-    # Note: ellmer will preserve full chat history automatically.
-    # this can confuse the agent and mess up behavior, so we reset:
-    chat$set_turns(NULL)
+    Always use the table 'gbif_aoi' rather than 'gbif' table if both are present.
 
-    } else {
-      output$agent <- renderText(response$agent)
-
-    }
+    IMPORTANT: return raw JSON only, do not decorate your reply with markdown code syntax.
+    ")
 
-  })
-  })
+    agent <- ellmer::chat_vllm(
+      base_url = "https://llm.cirrus.carlboettiger.info/v1/",
+      model = "kosbu/Llama-3.3-70B-Instruct-AWQ",
+      api_key = Sys.getenv("CIRRUS_LLM_KEY"),
+      system_prompt = system_prompt,
+      api_args = list(temperature = 0)
+    )
 
 
-  output$map <- renderMaplibre({
+    print("Agent thinking...")
+    stream <- agent$chat(input$chat)
 
-    m <- maplibre(center = c(-104.9, 40.3), zoom = 3, height = "400")
-    if (input$redlines) {
-      m <- m |>
-        add_fill_layer(
-          id = "redlines",
-          source = list(type = "vector",
-                        url = paste0("pmtiles://", "https://data.source.coop/cboettig/us-boundaries/mappinginequality.pmtiles")),
-          source_layer = "mappinginequality",
-          fill_color = list("get", "fill")
-        )
-    }
-    if (input$richness) {
-      m <- m |>
-        add_raster_source(id = "richness",
-                          tiles = "https://data.source.coop/cboettig/mobi/tiles/red/species-richness-all/{z}/{x}/{y}.png",
-                          maxzoom = 11
-                          ) |>
-        add_raster_layer(id = "richness-layer",
-                         source = "richness")
+    # Parse response
+    response <- jsonlite::fromJSON(stream)
 
-    }
+    if ("query" %in% names(response)) {
+      output$sql_code <- renderText(str_wrap(response$query, width = 60))
+      output$explanation <- renderText(response$explanation)
 
-     if (input$rsr) {
-      m <- m |>
-        add_raster_source(id = "rsr",
-                          tiles = "https://data.source.coop/cboettig/mobi/tiles/green/range-size-rarity-all/{z}/{x}/{y}.png",
-                          maxzoom = 11
-                          ) |>
-        add_raster_layer(id = "richness-layer",
-                         source = "rsr")
+    # clear agent memory
+    agent$set_turns(NULL)
 
+    } else {
+      output$agent <- renderText(response$agent)
     }
-    if (input$svi) {
-      m <- m |>
-        add_fill_layer(
-          id = "svi_layer",
-          source = list(type = "vector",
-                        url = paste0("pmtiles://", pmtiles)),
-          source_layer = "svi",
-          filter = filter_column(svi, data$df, "FIPS"),
-          fill_opacity = 0.5,
-          fill_color = interpolate(column = "RPL_THEMES",
-                                  values = c(0, 1),
-                                  stops = c("lightpink", "darkred"),
-                                  na_color = "lightgrey")
-        )
-    }
-  m |>
-    add_draw_control() |>
-    add_geocoder_control()
-
-  })
-
-  observeEvent(input$color, {
-    maplibre_proxy("map") |>
-      set_paint_property("svi_layer", "fill-color", input$color)
-  })
-  
-  observeEvent(input$slider, {
-    maplibre_proxy("map") |> 
-      set_filter("svi_layer", 
-                 list(">=", get_column("BIR74"), input$slider))
-  })
 
+    
+      # cache the query
+      query_id <- digest::digest(paste(response$query, bounds, collapse=""))
+      data_url <- glue::glue("https://minio.carlboettiger.info/public-data/cache/{query_id}.h3j")
+      output$url <- renderText(data_url)
+      
+      cache_parquet <- glue("{query_id}.parquet")
+
+      # compute if not yet in chache
+      status <- httr::status_code(httr::HEAD(data_url))
+      if(status == 404) {
+        print("Computing...")
+        time <- bench::bench_time({
+          agent_query(stream) |> write_dataset(cache_parquet)
+        })
+        print(time)
+      }
+
+      # draw on map
+      h3j <- glue("s3://public-data/cache/{query_id}.h3j")
+      open_dataset(cache_parquet) |> to_h3j(h3j)
+
+      # override previous map with drawn map
+      output$map <- renderMaplibre({
+          m <- maplibre(center=c(-110, 38), zoom = 3, pitch = 0, maxZoom = 9) |>
+          add_h3j_source("h3j_source",
+                        url = data_url) |>
+          add_fill_extrusion_layer(
+            id = "h3j_layer",
+            source = "h3j_source",
+            tooltip = "count",
+            fill_extrusion_color = interpolate(
+              column = "count",
+              values = c(0, 1000),
+              stops = c("#430254", "#f83c70")
+            ),
+            fill_extrusion_height = list(
+              "interpolate",
+              list("linear"),
+              list("zoom"),
+              0, 0, 1000,
+              list("*", 10, list("get", "count"))
+            ),
+            fill_extrusion_opacity = 0.7
+          )
+      }) # close renderMaplibre
+    }) # close observeEvent->get_features
+  }) # close observeEvent->user_msg
 }
 
 # Run the app

footer.md

@@ -1,16 +0,0 @@
-#### Credits
-
-Developed by Carl Boettiger, UC Berkeley, 2025.  BSD License.
-
-Data from the US Census and CDC's [Social Vulnerability Index](https://www.atsdr.cdc.gov/place-health/php/svi/index.html)
-
-#### Technical details
-
-The app is written entirely in R using shiny. The app will translate natural language queries in SQL code using
-a small open-weights language model. The SQL code is executed using the duckdb backend against cloud-native
-geoparquet snapshot of the Social Vulnerability Index hosted on Source Cooperative. Summary chart data are also
-computed in duckdb by streaming, providing responsive updates while needing minimal RAM or disk storage despite
-the large size of the data sources. 
-
-The map is rendered and updated using MapLibre with PMTiles, which provides responsive rendering for large feature sets.
-The PMTiles layer is also hosted on Source cooperative where it can be streamed efficiently.

system-prompt.md

@@ -1,34 +0,0 @@
-
-You are a helpful agent who always replies strictly in JSON-formatted text.
-Your task is to translate the user's questions about the data into a SQL query
-that will be run against the "biodiversity_occurrences" table in a duckdb database.
-The duckdb database has a spatial extension which understands PostGIS operations as well.
-
-If your answer involves the construction of a SQL query, you must format your answer as follows:
-
-{
-"query": "your raw SQL response goes here",
-"explanation": "your explanation of the query"
-}
-
-If your answer does not involve a SQL query, please reply with the following format instead:
-
-{
-    "user": "user question goes here",
-    "agent": "your response goes here"
-}
-
-If you are asked to describe the data or for information about the data schema, give only a human-readable response with SQL.
-
-In the data, each row represents an individual occurrence of a species. The occurrences
-are geocoded to US Census counties, with the STATE, COUNTY, and FIPS columns indicating
-the corresponding state name, county name, and FIPS identifier for the specific County.
-The FIPS column is an 5-digit number that uniquely identifies a county in a state.
-Taxonomic classification of the species is given in the corresponding columns, kingdom,
-phylum, class, order, family, genus, and species. 
-
-The data also includes information about various measures of social vulnerability (RPL_THEMES). 
-Pay attention to the DESCRIPTION of each of the columns (VARIABLE_NAME) from the metadata table:
-<schema>
-
-

test-app.R

@@ -1,26 +1,24 @@
-## Illustrate/test core app functionality without shiny
-
-library(tidyverse)
+library(duckdb.agent)
 library(duckdbfs)
-library(mapgl)
+library(dplyr)
 library(ellmer)
-library(glue)
 
-repo <- "https://data.source.coop/cboettig/social-vulnerability"
-pmtiles <- glue("{repo}/svi2020_us_tract.pmtiles")
-duckdb_s3_config(s3_endpoint = "minio.carlboettiger.info")
-svi <-
-  open_dataset("s3://public-gbif/svi", tblname = "biodiversity_occurrences") |>
-  filter(RPL_THEMES > 0)
-schema <- read_file("schema.yml")
-system_prompt <- glue::glue(readr::read_file("system-prompt.md"),
-                            .open = "<", .close = ">")
 
+duckdb_secrets(Sys.getenv("MINIO_KEY"),
+               Sys.getenv("MINIO_SECRET"),
+               "minio.carlboettiger.info")
+gbif <- open_dataset("s3://public-gbif/2024-10-01",  tblname = "gbif")
+tracts_url <- "https://minio.carlboettiger.info/public-social-vulnerability/2022-tracts-h3-z8.parquet"
+tracts_h3 <- open_dataset(tracts_url, tblname = "censustracts")
 
 
+system_prompt = create_prompt(additional_instructions = 
+"Note that the column h8 contains a geohash representing a H3 hexagon index.
+If asked for data that requires both tables, you should always seek to join on 
+the h8 column.  Always aggregate results to count the number of rows matching
+the query in each h8 hexagon")
 
-# Or optionally test with cirrus
-chat <- ellmer::chat_vllm(
+agent <- ellmer::chat_vllm(
   base_url = "https://llm.cirrus.carlboettiger.info/v1/",
   model = "kosbu/Llama-3.3-70B-Instruct-AWQ",
   api_key = Sys.getenv("CIRRUS_LLM_KEY"),
@@ -28,55 +26,38 @@ chat <- ellmer::chat_vllm(
   api_args = list(temperature = 0)
 )
 
-# or use the NRP model
-chat <- ellmer::chat_vllm(
-  base_url = "https://llm.nrp-nautilus.io/",
-  model = "olmo",
-  api_key = Sys.getenv("NRP_API_KEY"),
-  system_prompt = system_prompt,
-  api_args = list(temperature = 0)
-)
-
-cols <- colnames(svi)
-rpls <- grep("RPL_THEME.+", cols)
-keep <- cols[c(1:9, rpls, 161:226)]
-biodiversity <- svi |> select(all_of(keep))
-
-# Test a chat-based response
-chat$chat("Which columns describes racial components of social vulnerability?")
-chat$set_turns(NULL)
-## A query-based response
-stream <- chat$chat("Which counties have the most bird observations?")
-stream <- chat$chat("Give me the number bird observations per county vs county social vulnerability")
-response <- jsonlite::fromJSON(stream)
-
+resp <- agent$chat("Birds in Yolo County")
+out <- agent_query(resp)
 
-stream2 <- chat$chat("Great, now give me the ggplot2 code to plot the data.frame you returned as those counts vs social vulnerability as points.  Be sure to place the R code for your reply by itself in a 'code' element of the JSON")
+bench::bench_time({
+out |> rename(h3id = h6) |> to_h3j("s3://public-data/test5.h3j")
+})
 
-response <- jsonlite::fromJSON(stream2)
 
-con <- duckdbfs::cached_connection()
-filtered_data <- DBI::dbGetQuery(con, response$query)
 
-filter_column <- function(full_data, filtered_data, id_col) {
-  if (nrow(filtered_data) < 1) return(NULL)
-  values <- full_data |>
-    inner_join(filtered_data, copy = TRUE) |>
-    pull(id_col)
-  # maplibre syntax for the filter of PMTiles  
-  list("in", list("get", id_col), list("literal", values))
-}
-
-maplibre(center = c(-102.9, 41.3), zoom = 3) |>
-    add_fill_layer(
-        id = "svi_layer",
-        source = list(type = "vector", url  = paste0("pmtiles://", pmtiles)),
-        source_layer = "SVI2000_US_tract",
-        filter = filter_column(full_data, filtered_data, "FIPS"),
-        fill_opacity = 0.5,
-        fill_color = interpolate(column = "RPL_THEMES",
-                                values = c(0, 1),
-                                stops = c("#e19292c0", "darkblue"),
-                                na_color = "lightgrey")
-    )
+library(mapgl)
+url = "https://minio.carlboettiger.info/public-data/test5.h3j"
+maplibre(center=c(-110, 38), zoom = 3, pitch = 30) |>
+  add_h3j_source("h3j_source",
+                 url = url
+  )  |>
+  add_fill_extrusion_layer(
+    id = "h3j_layer",
+    source = "h3j_source",
+    fill_extrusion_color = interpolate(
+      column = "count",
+      values = c(0, 1000),
+      stops = c("#430254", "#f83c70")
+    ),
+    fill_extrusion_height = list(
+      "interpolate",
+      list("linear"),
+      list("zoom"),
+      0,
+      0,
+      100,
+      list("*", 2, list("get", "count"))
+    ),
+    fill_extrusion_opacity = 0.7
+  )
 

test-data.R

@@ -1,39 +1,23 @@
-library(dplyr)
-library(duckdbfs)
-library(mapgl)
 
-pad_tract = open_dataset("https://minio.carlboettiger.info/public-biodiversity/pad-us-4/pad-by-tract.parquet")
-pad = open_dataset("https://minio.carlboettiger.info/public-biodiversity/pad-us-4/pad-us-4.parquet")
-pmtiles = "https://minio.carlboettiger.info/public-biodiversity/pad-us-4/pad-us-4.pmtiles"
+library(duckdbfs)
+library(dplyr)
 
-#svi = open_dataset("https://data.source.coop/cboettig/social-vulnerability/2022/SVI2022_US_tract.parquet")
+library(sf)
+library(spData)
+bounds <- spData::us_states |>  dplyr::filter(NAME == "Arizona") |> sf::st_bbox()
+attach(as.list(bounds))
 
 
-State = "Colorado"
-filtered_data = pad_tract |> filter(STATE == State) |> select(Unit_Nm)
-full_data = pad |> filter(GAP_Sts %in% c("1", "2"))
+duckdb_secrets(Sys.getenv("MINIO_KEY"),
+               Sys.getenv("MINIO_SECRET"),
+               "minio.carlboettiger.info")
 
+gbif <- open_dataset("s3://public-gbif/2024-10-01",  tblname = "gbif")
+gbif_aoi <- 
+        gbif |> dplyr::filter(decimallatitude > 0) |> show_query()
 
 
 
-filter_column <- function(full_data, filtered_data, id_col) {
-  #if (nrow(filtered_data) < 1) return(NULL)
-  values <- full_data |>
-    inner_join(filtered_data, copy = TRUE) |>
-    pull(id_col)
-  # maplibre syntax for the filter of PMTiles  
-  list("in", list("get", id_col), list("literal", values))
-}
 
-maplibre(center = c(-102.9, 41.3), zoom = 3) |>
-    add_fill_layer(
-        id = "pad",
-        source = list(type = "vector", url  = paste0("pmtiles://", pmtiles)),
-        source_layer = "padus4",
-        tooltip = c("Unit_Nm"),
-        # filter = list("in", list("get", "GAP_Sts"), list("literal", values)),
-        filter = filter_column(full_data, filtered_data, "row_n"),
-        fill_opacity = 0.5,
-        fill_color = "darkgreen"
-    )
 
+gbif_aoi |> show_query()