Upload 3 files

Dockerfile

@@ -4,8 +4,10 @@
 # The `# syntax` line above enables BuildKit features (RUN --mount=type=secret).
 FROM rocker/geospatial:4.4
 
-# DuckDB lives on the edgarodriguez/depth_alpha dataset (public); baked in at
-# build. Override with: docker build --build-arg DDB_URL=... -t depth-hf .
+# DuckDB lives on the edgarodriguez/depth_alpha dataset (now PRIVATE); pulled with
+# the hf_token build secret below and baked in. Set that dataset to private (or move
+# the file to your private dataset and update this URL). Override with:
+#   docker build --secret id=hf_token,src=token.txt --build-arg DDB_URL=... -t depth-hf .
 ARG DDB_URL=https://huggingface.co/datasets/edgarodriguez/depth_alpha/resolve/main/depth_mexico.duckdb
 
 # RAG index (corpus chunks + local embeddings) for the AI assistant. The corpus
@@ -66,35 +68,42 @@ RUN mkdir -p "$OLLAMA_MODELS" \
 
 WORKDIR /app
 
-# Pre-bake the DuckDB vss extension under the runtime HOME (/app) so the
-# assistant's vector search works offline; the code falls back to core
-# list_cosine_distance if it is ever missing.
-RUN HOME=/app R -e "library(duckdb); con <- dbConnect(duckdb()); DBI::dbExecute(con, 'INSTALL vss; LOAD vss;'); DBI::dbDisconnect(con, shutdown=TRUE); cat('vss installed\n')"
+# Pre-bake the DuckDB vss (vector) + fts (BM25 keyword) extensions under the
+# runtime HOME (/app) so the assistant's hybrid retrieval works offline. The code
+# falls back to core list_cosine_distance / dense-only if either is missing.
+RUN HOME=/app R -e "library(duckdb); con <- dbConnect(duckdb()); DBI::dbExecute(con, 'INSTALL vss; LOAD vss; INSTALL fts; LOAD fts;'); DBI::dbDisconnect(con, shutdown=TRUE); cat('vss+fts installed\n')"
 
-COPY depth_interactive_map_alpha.R assistant.R helpers.R styles.css entrypoint.sh ./
+# indicator_dictionary.json (generated by R/96) grounds the assistant's tools +
+# alias routing; assistant.R loads it from /app/. Regenerate + re-copy when the
+# data dictionary overlay changes.
+COPY depth_interactive_map_alpha.R assistant.R helpers.R styles.css entrypoint.sh indicator_dictionary.json ./
 
-# Bake the DuckDB map data (public dataset, build-time download).
-RUN curl -fL "${DDB_URL}" -o /app/depth_mexico.duckdb \
- && test -s /app/depth_mexico.duckdb \
- && ls -lh /app/depth_mexico.duckdb
-
-# Bake the private RAG index via the hf_token secret (best-effort; never fails
-# the build). The secret is mounted only for this RUN, at /run/secrets/hf_token.
-# Reports the HTTP status on failure (401 = bad/expired token, 403 = token has no
-# read access to the dataset, 404 = wrong URL/path) and removes a partial file so
-# the app correctly shows "offline" rather than loading a broken index.
+# Bake the PRIVATE datasets via the hf_token build secret (mounted only for this
+# RUN at /run/secrets/hf_token). The map DuckDB is MANDATORY -- the app cannot run
+# without it, so a non-200 FAILS the build with the HTTP code (401 = bad/expired
+# token, 403 = token has no read access, 404 = wrong URL/path). The RAG index is
+# best-effort (a partial file is removed so the assistant shows "offline" rather
+# than loading a broken index).
 RUN --mount=type=secret,id=hf_token \
-    if [ -s /run/secrets/hf_token ]; then \
-      code=$(curl -sL -o /app/rag_chunks.parquet -w '%{http_code}' \
-             -H "Authorization: Bearer $(cat /run/secrets/hf_token)" "${RAG_URL}"); \
-      if [ "$code" = "200" ] && [ -s /app/rag_chunks.parquet ]; then \
-        echo "[build] RAG index baked: $(ls -lh /app/rag_chunks.parquet | awk '{print $5}')"; \
-      else \
-        echo "[build] RAG fetch failed (HTTP $code) -> assistant offline"; \
-        rm -f /app/rag_chunks.parquet; \
-      fi; \
+    if [ ! -s /run/secrets/hf_token ]; then \
+      echo "[build] ERROR: hf_token secret missing -- add it as a Space secret"; exit 1; \
+    fi; \
+    TOKEN="$(cat /run/secrets/hf_token)"; \
+    dcode=$(curl -sL -o /app/depth_mexico.duckdb -w '%{http_code}' \
+            -H "Authorization: Bearer $TOKEN" "${DDB_URL}"); \
+    if [ "$dcode" = "200" ] && [ -s /app/depth_mexico.duckdb ]; then \
+      echo "[build] DuckDB baked: $(ls -lh /app/depth_mexico.duckdb | awk '{print $5}')"; \
+    else \
+      echo "[build] ERROR: DuckDB fetch failed (HTTP $dcode) -- check token / DDB_URL / dataset visibility"; \
+      rm -f /app/depth_mexico.duckdb; exit 1; \
+    fi; \
+    rcode=$(curl -sL -o /app/rag_chunks.parquet -w '%{http_code}' \
+            -H "Authorization: Bearer $TOKEN" "${RAG_URL}"); \
+    if [ "$rcode" = "200" ] && [ -s /app/rag_chunks.parquet ]; then \
+      echo "[build] RAG index baked: $(ls -lh /app/rag_chunks.parquet | awk '{print $5}')"; \
     else \
-      echo "[build] no hf_token secret found at /run/secrets/hf_token -> assistant offline"; \
+      echo "[build] RAG fetch failed (HTTP $rcode) -> assistant offline"; \
+      rm -f /app/rag_chunks.parquet; \
     fi
 
 # HF Spaces runs the container as a non-root user -- make /app world-readable,

assistant.R

@@ -60,53 +60,100 @@ ollama_embed <- function(text, cfg = assistant_config()) {
   as.numeric(unlist(out$embeddings[[1]]))
 }
 
-# Decide once whether the vss extension (array_cosine_distance) is available;
-# fall back to DuckDB's core list_cosine_distance otherwise. Both are exact
-# brute-force cosine -- fine for a few thousand chunks.
 .assist_cache <- new.env(parent = emptyenv())
 
-.rag_dist_clause <- function(con, vec_lit, d) {
-  mode <- .assist_cache$dist_mode
-  if (is.null(mode)) {
-    ok <- tryCatch({
-      DBI::dbExecute(con, "INSTALL vss; LOAD vss;")
-      DBI::dbGetQuery(con,
-        "SELECT array_cosine_distance(CAST([1.0,2.0] AS FLOAT[2]), CAST([1.0,2.0] AS FLOAT[2])) AS d")
-      TRUE
-    }, error = function(e) FALSE)
-    mode <- if (isTRUE(ok)) "vss" else "list"
-    .assist_cache$dist_mode <- mode
-    message("[assist] vector search mode: ", mode)
-  }
-  if (mode == "vss") {
+# Model-aware query text: nomic-embed-text needs the "search_query: " task prefix;
+# bge-m3 and most other embedders do not.
+.embed_query_text <- function(query, cfg) {
+  if (grepl("nomic", cfg$embed_model, ignore.case = TRUE)) paste0("search_query: ", query)
+  else query
+}
+
+# Build (once per process) a writable in-memory DuckDB holding the rag chunks, with
+# the vss (vector) extension and an fts (BM25) index for hybrid search. Returns NULL
+# if the parquet is missing or it can't be built. Cached in .assist_cache.
+.get_rag_con <- function(cfg) {
+  if (!is.null(.assist_cache$rag_con)) return(.assist_cache$rag_con)
+  if (!file.exists(cfg$rag_path)) return(NULL)
+  rc <- tryCatch({
+    c2 <- DBI::dbConnect(duckdb::duckdb())
+    tryCatch(DBI::dbExecute(c2, "INSTALL vss; LOAD vss;"), error = function(e) NULL)
+    DBI::dbExecute(c2, sprintf("CREATE TABLE rag AS SELECT * FROM read_parquet(%s)",
+                               DBI::dbQuoteString(c2, cfg$rag_path)))
+    c2
+  }, error = function(e) { message("[assist] rag_con: ", conditionMessage(e)); NULL })
+  if (is.null(rc)) return(NULL)
+  .assist_cache$vss_ok <- tryCatch({
+    DBI::dbGetQuery(rc, "SELECT array_cosine_distance(CAST([1.0] AS FLOAT[1]), CAST([1.0] AS FLOAT[1]))")
+    TRUE
+  }, error = function(e) FALSE)
+  .assist_cache$fts_ok <- tryCatch({
+    DBI::dbExecute(rc, "INSTALL fts; LOAD fts;")
+    DBI::dbExecute(rc,
+      "PRAGMA create_fts_index('rag','chunk_id','chunk_text', stemmer='none', strip_accents=1, lower=1, overwrite=1)")
+    TRUE
+  }, error = function(e) { message("[assist] fts: ", conditionMessage(e)); FALSE })
+  message("[assist] rag index ready (vss=", .assist_cache$vss_ok,
+          " fts=", .assist_cache$fts_ok, ")")
+  .assist_cache$rag_con <- rc
+  rc
+}
+
+.dist_clause <- function(vss_ok, vec_lit, d) {
+  if (isTRUE(vss_ok))
     glue::glue("array_cosine_distance(CAST(embedding AS FLOAT[{d}]), CAST({vec_lit} AS FLOAT[{d}]))")
-  } else {
+  else
     glue::glue("list_cosine_distance(embedding, CAST({vec_lit} AS DOUBLE[]))")
-  }
 }
 
-# Embed the question (with the nomic query prefix) and return the top-k corpus
-# passages as a single context string with [source] tags for citation. k kept
-# small (2) so the injected context stays short -> cheaper CPU prefill.
-rag_search <- function(con, query, cfg = assistant_config(), k = 2) {
-  if (is.null(con) || !file.exists(cfg$rag_path)) return("")
-  qvec <- ollama_embed(paste0("search_query: ", query), cfg)
+# Hybrid retrieval: dense cosine + BM25 keyword, fused with Reciprocal Rank Fusion,
+# returning the top-k passages as one context string with [source] tags. Degrades to
+# dense-only if fts is unavailable. k kept small so injected context stays cheap on
+# CPU prefill. `con` is unused (kept for signature stability).
+rag_search <- function(con, query, cfg = assistant_config(), k = 2, n_cand = 20) {
+  rc <- .get_rag_con(cfg)
+  if (is.null(rc)) return("")
+  qvec <- ollama_embed(.embed_query_text(query, cfg), cfg)
   if (length(qvec) == 0) return("")
   vec_lit <- paste0("[", paste(format(qvec, scientific = FALSE, trim = TRUE),
                                collapse = ","), "]")
-  dist <- .rag_dist_clause(con, vec_lit, length(qvec))
-  sql <- glue::glue(
-    "SELECT source_title, folder, chunk_text, {dist} AS dist
-     FROM read_parquet('{path}')
-     ORDER BY dist ASC
-     LIMIT {k}",
-    path = cfg$rag_path, k = as.integer(k)
-  )
-  res <- tryCatch(DBI::dbGetQuery(con, sql), error = function(e) {
-    message("[assist] rag_search: ", conditionMessage(e)); NULL
-  })
-  if (is.null(res) || nrow(res) == 0) return("")
-  paste(sprintf("[%s] %s", res$source_title, res$chunk_text), collapse = "\n\n")
+  dist <- .dist_clause(.assist_cache$vss_ok, vec_lit, length(qvec))
+
+  dense <- tryCatch(DBI::dbGetQuery(rc, glue::glue(
+    "SELECT chunk_id, {dist} AS dist FROM rag ORDER BY dist ASC LIMIT {n_cand}")),
+    error = function(e) { message("[assist] dense: ", conditionMessage(e)); NULL })
+  if (is.null(dense) || nrow(dense) == 0) return("")
+  dense$rank <- seq_len(nrow(dense))
+
+  sparse <- NULL
+  if (isTRUE(.assist_cache$fts_ok)) {
+    qx <- DBI::dbQuoteString(rc, query)
+    sparse <- tryCatch(DBI::dbGetQuery(rc, glue::glue(
+      "SELECT chunk_id, score FROM (
+         SELECT chunk_id, fts_main_rag.match_bm25(chunk_id, {qx}) AS score FROM rag
+       ) WHERE score IS NOT NULL ORDER BY score DESC LIMIT {n_cand}")),
+      error = function(e) { message("[assist] bm25: ", conditionMessage(e)); NULL })
+  }
+
+  # Reciprocal Rank Fusion (k_const = 60).
+  scores <- setNames(1 / (60 + dense$rank), as.character(dense$chunk_id))
+  if (!is.null(sparse) && nrow(sparse) > 0) {
+    sr <- 1 / (60 + seq_len(nrow(sparse)))
+    for (i in seq_len(nrow(sparse))) {
+      id <- as.character(sparse$chunk_id[i])
+      scores[id] <- (if (is.na(scores[id])) 0 else scores[id]) + sr[i]
+    }
+  }
+  top_ids <- as.integer(names(sort(scores, decreasing = TRUE)))[seq_len(min(k, length(scores)))]
+  top_ids <- top_ids[!is.na(top_ids)]
+  if (length(top_ids) == 0) return("")
+  txt <- tryCatch(DBI::dbGetQuery(rc, glue::glue(
+    "SELECT chunk_id, source_title, chunk_text FROM rag WHERE chunk_id IN ({paste(top_ids, collapse = ',')})")),
+    error = function(e) NULL)
+  if (is.null(txt) || nrow(txt) == 0) return("")
+  txt <- txt[match(top_ids, txt$chunk_id), ]   # preserve fused order
+  txt <- txt[!is.na(txt$chunk_id), ]
+  paste(sprintf("[%s] %s", txt$source_title, txt$chunk_text), collapse = "\n\n")
 }
 
 # --- Grounding text from WEB_SPEC --------------------------------------------
@@ -156,18 +203,94 @@ assistant_user_turn <- function(question, context_txt = "", scope_txt = "") {
 # `state` is a non-reactive environment the server keeps in sync with rv, so
 # tools can be called outside a reactive context during async streaming.
 
-.IND_LABELS <- c(
-  homicidio       = "homicide rate (per 100k)",
-  robo            = "robbery rate (per 100k)",
-  secuestro       = "kidnapping rate (per 100k)",
-  trafico_menores = "child-trafficking rate (per 100k)",
-  trata_personas  = "human-trafficking rate (per 100k)",
-  desap_n         = "disappeared and not located (count)",
-  pam_n           = "irregular-migration (PAM) events (count)"
+# Queryable municipality-level indicators. Each maps to a DuckDB table + value
+# expression, joined by CVEGEO. `per_muni=TRUE` means the table has many rows per
+# municipality (sum them first); `kind` sets the sensible default statistic
+# (rate -> mean across munis, count -> sum). All join to crime_mun for names.
+.IND_SPEC <- list(
+  homicidio       = list(table = "crime_mun",       expr = "CAST(homicidio AS DOUBLE)",        per_muni = FALSE, kind = "rate",  label = "homicide rate",                 unit = "per 100k"),
+  robo            = list(table = "crime_mun",       expr = "CAST(robo AS DOUBLE)",             per_muni = FALSE, kind = "rate",  label = "robbery rate",                  unit = "per 100k"),
+  secuestro       = list(table = "crime_mun",       expr = "CAST(secuestro AS DOUBLE)",        per_muni = FALSE, kind = "rate",  label = "kidnapping rate",               unit = "per 100k"),
+  lesiones        = list(table = "crime_mun",       expr = "CAST(lesiones AS DOUBLE)",         per_muni = FALSE, kind = "rate",  label = "intentional-injury rate",       unit = "per 100k"),
+  trafico_menores = list(table = "crime_mun",       expr = "CAST(trafico_menores AS DOUBLE)",  per_muni = FALSE, kind = "rate",  label = "child-trafficking rate",        unit = "per 100k"),
+  trata_personas  = list(table = "crime_mun",       expr = "CAST(trata_personas AS DOUBLE)",   per_muni = FALSE, kind = "rate",  label = "human-trafficking rate",        unit = "per 100k"),
+  poblacion       = list(table = "crime_mun",       expr = "CAST(poblacion AS DOUBLE)",        per_muni = FALSE, kind = "count", label = "population",                     unit = "people"),
+  desap_n         = list(table = "dim_desap_mun",   expr = "final_desap_nl",                   per_muni = TRUE,  kind = "count", label = "disappeared and not located",   unit = "count"),
+  pam_n           = list(table = "dim_pam_mun",     expr = "pam_n",                            per_muni = TRUE,  kind = "count", label = "irregular-migration (PAM) events", unit = "count"),
+  fosas_total     = list(table = "dim_fosas_mun",   expr = "fosas_total",                      per_muni = TRUE,  kind = "count", label = "clandestine graves",            unit = "count"),
+  cuerpos_total   = list(table = "dim_fosas_mun",   expr = "cuerpos_total",                    per_muni = TRUE,  kind = "count", label = "bodies recovered",              unit = "count"),
+  pobreza_pct     = list(table = "dim_pobreza_mun", expr = "TRY_CAST(pobreza_pct AS DOUBLE)",  per_muni = FALSE, kind = "rate",  label = "poverty",                       unit = "%"),
+  pobreza_ext_pct = list(table = "dim_pobreza_mun", expr = "TRY_CAST(pobreza_ext_pct AS DOUBLE)", per_muni = FALSE, kind = "rate", label = "extreme poverty",            unit = "%"),
+  remesas_2024    = list(table = "dim_remesas_mun", expr = "TRY_CAST(remesas_2024 AS DOUBLE)", per_muni = FALSE, kind = "count", label = "remittances 2024",              unit = "MXN")
 )
 .POINT_TO_COUNT <- c(migrants = "n_incidents", graves = "n_graves",
                      infra = "n_facilities", ocved = "ocved_n", ged = "n_ged")
 
+# Optional enrichment: load the generated indicator_dictionary.json (labels, units,
+# aliases) to refine tool descriptions + alias routing. Absent -> use .IND_SPEC.
+.load_ind_dict <- function() {
+  if (!is.null(.assist_cache$ind_dict)) return(.assist_cache$ind_dict)
+  cands <- c(Sys.getenv("DEPTH_INDICATOR_DICT", ""), "/app/indicator_dictionary.json",
+             "data/export/web/indicator_dictionary.json")
+  cands <- cands[nzchar(cands) & file.exists(cands)]
+  d <- if (length(cands) > 0)
+    tryCatch(jsonlite::fromJSON(cands[1], simplifyVector = FALSE)$indicators,
+             error = function(e) NULL) else NULL
+  .assist_cache$ind_dict <- d %||% list()
+  .assist_cache$ind_dict
+}
+
+# alias (and label) -> canonical .IND_SPEC key, so "asesinatos"/"murders" -> homicidio.
+.alias_map <- function() {
+  if (!is.null(.assist_cache$alias_map)) return(.assist_cache$alias_map)
+  m <- list()
+  for (k in names(.IND_SPEC)) m[[tolower(k)]] <- k
+  for (it in .load_ind_dict()) {
+    key <- it$key
+    if (is.null(key) || is.null(.IND_SPEC[[key]])) next
+    for (a in unlist(it$aliases)) if (nzchar(a)) m[[tolower(trimws(a))]] <- key
+    if (nzchar(it$label_en %||% "")) m[[tolower(it$label_en)]] <- key
+  }
+  .assist_cache$alias_map <- m
+  m
+}
+
+.resolve_indicator <- function(x) {
+  if (is.null(x) || !nzchar(x)) return(NA_character_)
+  if (!is.null(.IND_SPEC[[x]])) return(x)
+  am <- .alias_map(); xl <- tolower(trimws(x))
+  if (!is.null(am[[xl]])) return(am[[xl]])
+  NA_character_
+}
+
+# Human label "<label> (<unit>)", preferring the dictionary's wording.
+.ind_label <- function(key) {
+  spec <- .IND_SPEC[[key]]
+  de <- Filter(function(it) identical(it$key, key), .load_ind_dict())
+  lab  <- if (length(de) > 0 && nzchar(de[[1]]$label_en %||% "")) de[[1]]$label_en else spec$label
+  unit <- if (length(de) > 0 && nzchar(de[[1]]$unit %||% ""))     de[[1]]$unit     else spec$unit
+  if (nzchar(unit)) paste0(lab, " (", unit, ")") else lab
+}
+
+# CVEGEO list of the current selection (NULL when national / nothing selected).
+.cvegeo_filter <- function(state) {
+  if (isTRUE(state$is_national) || is.null(state$mun_df) ||
+      is.null(state$mun_df$CVEGEO) || nrow(state$mun_df) == 0) return(NULL)
+  cv <- unique(state$mun_df$CVEGEO)
+  cv <- cv[!is.na(cv) & nzchar(cv)]
+  if (length(cv) == 0) NULL else cv
+}
+
+# Per-municipality (CVEGEO, value) subquery for an indicator spec.
+.base_subquery <- function(spec) {
+  if (isTRUE(spec$per_muni))
+    glue::glue("(SELECT CVEGEO AS k, SUM({spec$expr}) AS v FROM {spec$table} GROUP BY CVEGEO)")
+  else
+    glue::glue("(SELECT CVEGEO AS k, {spec$expr} AS v FROM {spec$table})")
+}
+
+.quote_cvegeos <- function(cv) paste0("'", gsub("'", "''", cv), "'", collapse = ",")
+
 assistant_selection_summary <- function(state) {
   cnt <- state$counts %||% list()
   scope <- if (isTRUE(state$is_national)) {
@@ -200,83 +323,51 @@ assistant_count_points <- function(state, layer) {
 }
 
 assistant_aggregate <- function(con, state, indicator, statistic = "mean") {
-  if (!indicator %in% names(.IND_LABELS)) return(paste0("Unknown indicator: ", indicator))
-  if (!statistic %in% c("mean", "sum", "max", "min")) statistic <- "mean"
-  lab <- .IND_LABELS[[indicator]]
-
-  # Selection path: aggregate the in-memory municipalities (matches the sidebar).
-  if (!isTRUE(state$is_national) && !is.null(state$mun_df) &&
-      nrow(state$mun_df) > 0 && indicator %in% names(state$mun_df)) {
-    vals <- suppressWarnings(as.numeric(state$mun_df[[indicator]]))
-    vals <- vals[is.finite(vals)]
-    if (length(vals) == 0) return(glue::glue("No {lab} values in the current selection."))
-    res <- switch(statistic, mean = mean(vals), sum = sum(vals),
-                  max = max(vals), min = min(vals))
-    return(glue::glue(
-      "{statistic} of {lab} across {nrow(state$mun_df)} selected municipalities: {round(res, 2)}."))
-  }
-
-  # National path: a single bounded aggregate query (indicator is allow-listed).
+  key <- .resolve_indicator(indicator)
+  if (is.na(key)) return(paste0("Unknown indicator: ", indicator,
+    ". Available: ", paste(names(.IND_SPEC), collapse = ", ")))
   if (is.null(con)) return("No database connection available.")
-  if (indicator %in% c("desap_n", "pam_n")) {
-    inner_tbl <- if (indicator == "desap_n") "dim_desap_mun" else "dim_pam_mun"
-    inner_col <- if (indicator == "desap_n") "final_desap_nl" else "pam_n"
-    from_clause <- glue::glue("(SELECT CVEGEO, SUM({inner_col}) AS x FROM {inner_tbl} GROUP BY CVEGEO)")
-    col_expr <- "x"
-  } else {
-    from_clause <- "crime_mun"
-    col_expr <- glue::glue("CAST({indicator} AS DOUBLE)")
-  }
-  agg <- switch(statistic,
-    mean = glue::glue("AVG({col_expr})"), sum = glue::glue("SUM({col_expr})"),
-    max  = glue::glue("MAX({col_expr})"), min = glue::glue("MIN({col_expr})"))
-  sql <- glue::glue("SELECT {agg} AS v, COUNT(*) AS n FROM {from_clause}")
-  r <- tryCatch(DBI::dbGetQuery(con, sql), error = function(e) NULL)
-  if (is.null(r) || nrow(r) == 0 || is.na(r$v[1]))
-    return(glue::glue("Could not compute {lab} nationally."))
-  glue::glue("{statistic} of {lab} across all {r$n[1]} municipalities in Mexico: {round(r$v[1], 2)}.")
+  spec <- .IND_SPEC[[key]]
+  if (!statistic %in% c("mean", "sum", "max", "min"))
+    statistic <- if (spec$kind == "rate") "mean" else "sum"
+  lab  <- .ind_label(key)
+  base <- .base_subquery(spec)
+  cv   <- .cvegeo_filter(state)
+  where <- if (!is.null(cv)) glue::glue("WHERE k IN ({.quote_cvegeos(cv)})") else ""
+  agg <- switch(statistic, mean = "AVG(v)", sum = "SUM(v)", max = "MAX(v)", min = "MIN(v)")
+  sql <- glue::glue("SELECT {agg} AS res, COUNT(v) AS n FROM {base} {where}")
+  r <- tryCatch(DBI::dbGetQuery(con, sql),
+                error = function(e) { message("[assist] agg: ", conditionMessage(e)); NULL })
+  if (is.null(r) || nrow(r) == 0 || is.na(r$res[1]))
+    return(glue::glue("No {lab} data for that scope."))
+  scope <- if (!is.null(cv)) glue::glue("{r$n[1]} selected municipalities")
+           else glue::glue("all {r$n[1]} municipalities in Mexico")
+  glue::glue("{statistic} of {lab} across {scope}: {round(r$res[1], 2)}.")
 }
 
 assistant_top_municipalities <- function(con, state, indicator, n = 5) {
-  if (!indicator %in% names(.IND_LABELS)) return(paste0("Unknown indicator: ", indicator))
-  n   <- max(1L, min(20L, as.integer(n)))
-  lab <- .IND_LABELS[[indicator]]
-
-  # Selection path: rank the in-memory municipalities.
-  if (!isTRUE(state$is_national) && !is.null(state$mun_df) &&
-      nrow(state$mun_df) > 0 && indicator %in% names(state$mun_df)) {
-    d <- state$mun_df
-    d$.v <- suppressWarnings(as.numeric(d[[indicator]]))
-    d <- d[is.finite(d$.v), , drop = FALSE]
-    if (nrow(d) == 0) return(glue::glue("No {lab} data in the current selection."))
-    d <- utils::head(d[order(-d$.v), , drop = FALSE], n)
-    items <- sprintf("%d. %s (%s): %s", seq_len(nrow(d)), d$municipio, d$entidad,
-                     round(d$.v, 2))
-    return(paste0("Top municipalities by ", lab, " (current selection):\n",
-                  paste(items, collapse = "\n")))
-  }
-
-  # National path.
+  key <- .resolve_indicator(indicator)
+  if (is.na(key)) return(paste0("Unknown indicator: ", indicator,
+    ". Available: ", paste(names(.IND_SPEC), collapse = ", ")))
   if (is.null(con)) return("No database connection available.")
-  if (indicator %in% c("homicidio", "robo", "secuestro", "trafico_menores", "trata_personas")) {
-    sql <- glue::glue(
-      "SELECT municipio, entidad, CAST({indicator} AS DOUBLE) AS v
-       FROM crime_mun WHERE {indicator} IS NOT NULL
-       ORDER BY v DESC NULLS LAST LIMIT {n}")
-  } else {
-    inner_tbl <- if (indicator == "desap_n") "dim_desap_mun" else "dim_pam_mun"
-    inner_col <- if (indicator == "desap_n") "final_desap_nl" else "pam_n"
-    sql <- glue::glue(
-      "SELECT c.municipio, c.entidad, s.v AS v
-       FROM crime_mun c
-       JOIN (SELECT CVEGEO, SUM({inner_col}) AS v FROM {inner_tbl} GROUP BY CVEGEO) s
-         ON c.CVEGEO = s.CVEGEO
-       ORDER BY v DESC NULLS LAST LIMIT {n}")
-  }
-  r <- tryCatch(DBI::dbGetQuery(con, sql), error = function(e) NULL)
-  if (is.null(r) || nrow(r) == 0) return(glue::glue("Could not rank municipalities by {lab}."))
+  n <- suppressWarnings(as.integer(n)); if (is.na(n)) n <- 5L
+  n <- max(1L, min(20L, n))
+  spec <- .IND_SPEC[[key]]; lab <- .ind_label(key)
+  base <- .base_subquery(spec)
+  cv <- .cvegeo_filter(state)
+  conds <- "b.v IS NOT NULL"
+  if (!is.null(cv)) conds <- paste0(conds, " AND b.k IN (", .quote_cvegeos(cv), ")")
+  sql <- glue::glue(
+    "SELECT c.municipio, c.entidad, b.v AS v
+     FROM {base} b JOIN crime_mun c ON c.CVEGEO = b.k
+     WHERE {conds}
+     ORDER BY v DESC NULLS LAST LIMIT {n}")
+  r <- tryCatch(DBI::dbGetQuery(con, sql),
+                error = function(e) { message("[assist] top: ", conditionMessage(e)); NULL })
+  if (is.null(r) || nrow(r) == 0) return(glue::glue("No {lab} data for that scope."))
+  scope <- if (!is.null(cv)) "current selection" else "Mexico"
   items <- sprintf("%d. %s (%s): %s", seq_len(nrow(r)), r$municipio, r$entidad, round(r$v, 2))
-  paste0("Top municipalities by ", lab, " (Mexico):\n", paste(items, collapse = "\n"))
+  paste0("Top municipalities by ", lab, " (", scope, "):\n", paste(items, collapse = "\n"))
 }
 
 # --- Assemble the per-session chat -------------------------------------------
@@ -300,7 +391,10 @@ make_assistant <- function(con, state, cfg = assistant_config()) {
   )
   if (is.null(chat)) return(NULL)
 
-  ind_enum <- ellmer::type_enum(names(.IND_LABELS), "Indicator column.")
+  ind_desc <- paste0("Indicator key. Options -- ",
+    paste(vapply(names(.IND_SPEC), \(k) paste0(k, " (", .ind_label(k), ")"), character(1)),
+          collapse = "; "))
+  ind_enum <- ellmer::type_enum(names(.IND_SPEC), ind_desc)
 
   # Only the 3 data tools are registered. The indicator catalog lives in the
   # system prompt and the current scope/totals are injected as <scope> each turn,
@@ -313,7 +407,7 @@ make_assistant <- function(con, state, cfg = assistant_config()) {
   count_points <- function(layer) assistant_count_points(state, layer)
 
   chat$register_tool(ellmer::tool(aggregate_indicator,
-    "Aggregate one municipality indicator over the current selection (or nationally if nothing is selected). Use 'mean' for rate indicators and 'sum' for count indicators.",
+    "Aggregate one municipality indicator over the current map selection (or nationally if nothing is selected). Use 'mean' for rate indicators (rates, %) and 'sum' for counts.",
     arguments = list(
       indicator = ind_enum,
       statistic = ellmer::type_enum(c("mean", "sum", "max", "min"),

indicator_dictionary.json

@@ -0,0 +1,1415 @@
+{
+  "generated": "2026-06-22",
+  "source": "data/export/duckdb/depth_mexico.duckdb",
+  "n_indicators": 128,
+  "indicators": [
+    {
+      "key": "poblacion",
+      "table": "crime_mun",
+      "type": "DOUBLE",
+      "label_en": "Population",
+      "label_es": "Población",
+      "unit": "people",
+      "source": "SESNSP (Sistema Nacional de Seguridad Publica), 2022-2025",
+      "range": [282, 7351266],
+      "aliases": ["population", "poblacion", "habitantes"]
+    },
+    {
+      "key": "homicidio",
+      "table": "crime_mun",
+      "type": "DOUBLE",
+      "label_en": "Homicides",
+      "label_es": "Tasa de homicidios dolosos",
+      "unit": "per 100k",
+      "source": "SESNSP (Sistema Nacional de Seguridad Publica), 2022-2025",
+      "range": [0, 783.73],
+      "aliases": ["homicide", "murders", "homicidios", "asesinatos"]
+    },
+    {
+      "key": "lesiones",
+      "table": "crime_mun",
+      "type": "DOUBLE",
+      "label_en": "Intentional injury rate",
+      "label_es": "Tasa de lesiones dolosas",
+      "unit": "per 100k",
+      "source": "SESNSP (Sistema Nacional de Seguridad Publica), 2022-2025",
+      "range": [0, 768.27],
+      "aliases": ["injuries", "assault", "lesiones"]
+    },
+    {
+      "key": "rapto",
+      "table": "crime_mun",
+      "type": "DOUBLE",
+      "label_en": "Abduction (non-extortion) rate",
+      "label_es": "Tasa de rapto",
+      "unit": "per 100k",
+      "source": "SESNSP (Sistema Nacional de Seguridad Publica), 2022-2025",
+      "range": [0, 9.63],
+      "aliases": ["abduction", "rapto"]
+    },
+    {
+      "key": "robo",
+      "table": "crime_mun",
+      "type": "DOUBLE",
+      "label_en": "Robberies",
+      "label_es": "Tasa de robo",
+      "unit": "per 100k",
+      "source": "SESNSP (Sistema Nacional de Seguridad Publica), 2022-2025",
+      "range": [0, 2189.24],
+      "aliases": ["robbery", "theft", "robo", "asalto"]
+    },
+    {
+      "key": "secuestro",
+      "table": "crime_mun",
+      "type": "DOUBLE",
+      "label_en": "Kidnapping",
+      "label_es": "Tasa de secuestro",
+      "unit": "per 100k",
+      "source": "SESNSP (Sistema Nacional de Seguridad Publica), 2022-2025",
+      "range": [0, 26.47],
+      "aliases": ["kidnapping", "secuestro", "abduction"]
+    },
+    {
+      "key": "trafico_menores",
+      "table": "crime_mun",
+      "type": "DOUBLE",
+      "label_en": "child trafficking",
+      "label_es": "Tasa de trafico de menores",
+      "unit": "per 100k",
+      "source": "SESNSP (Sistema Nacional de Seguridad Publica), 2022-2025",
+      "range": [0, 0.86],
+      "aliases": ["child trafficking", "trafico de menores", "menores"]
+    },
+    {
+      "key": "trata_personas",
+      "table": "crime_mun",
+      "type": "DOUBLE",
+      "label_en": "human trafficking",
+      "label_es": "Tasa de trata de personas",
+      "unit": "per 100k",
+      "source": "SESNSP (Sistema Nacional de Seguridad Publica), 2022-2025",
+      "range": [0, 26.09],
+      "aliases": ["human trafficking", "trata", "trafficking"]
+    },
+    {
+      "key": "final_desap_nl",
+      "table": "dim_desap_mun",
+      "type": "DOUBLE",
+      "label_en": "disappearance and not located people",
+      "label_es": "Desaparecidos y no localizados",
+      "unit": "count",
+      "source": "RNPDNO (Registro Nacional de Personas Desaparecidas), 2014-2023",
+      "range": [1.0001, 1827.1838],
+      "aliases": ["disappearances", "missing", "desaparecidos", "no localizados"]
+    },
+    {
+      "key": "final_desap_nl_per_100k",
+      "table": "dim_desap_mun",
+      "type": "DOUBLE",
+      "label_en": "disappearance and not located per 100k inhabitants",
+      "label_es": "Desaparecidos no localizados por 100 mil habitantes",
+      "unit": "per 100k",
+      "source": "RNPDNO (Registro Nacional de Personas Desaparecidas), 2014-2023",
+      "range": [0.051, 5653.7652],
+      "aliases": ["disappearances rate", "desaparecidos por 100k"]
+    },
+    {
+      "key": "dark_figures_perc",
+      "table": "dim_envipe_state",
+      "type": "DOUBLE",
+      "label_en": "Dark figures (unreported crime)",
+      "label_es": "Cifra negra (delitos no denunciados)",
+      "unit": "%",
+      "source": "ENVIPE (Encuesta Nacional de Victimizacion y Percepcion sobre Seguridad Publica), 2022-2023",
+      "range": [88.2671, 97.1171],
+      "aliases": ["dark figure", "unreported crime", "cifra negra"]
+    },
+    {
+      "key": "perception_insecurity_perc",
+      "table": "dim_envipe_state",
+      "type": "DOUBLE",
+      "label_en": "Insecurity perspection",
+      "label_es": "Percepcion de inseguridad",
+      "unit": "%",
+      "source": "ENVIPE (Encuesta Nacional de Victimizacion y Percepcion sobre Seguridad Publica), 2022-2023",
+      "range": [20.419, 57.993],
+      "aliases": ["insecurity perception", "percepcion de inseguridad", "inseguridad"]
+    },
+    {
+      "key": "year",
+      "table": "dim_fosas_mun",
+      "type": "INTEGER",
+      "label_en": "Year",
+      "label_es": "Año",
+      "unit": "",
+      "source": "Plataforma Ciudadana de Fosas, 2006-2024",
+      "range": [2007, 2024],
+      "aliases": []
+    },
+    {
+      "key": "fosas_total",
+      "table": "dim_fosas_mun",
+      "type": "DOUBLE",
+      "label_en": "Clandestine graves",
+      "label_es": "Fosas clandestinas",
+      "unit": "count",
+      "source": "Plataforma Ciudadana de Fosas, 2006-2024",
+      "range": [1, 102.5],
+      "aliases": ["graves", "clandestine graves", "fosas", "fosas clandestinas"]
+    },
+    {
+      "key": "cuerpos_total",
+      "table": "dim_fosas_mun",
+      "type": "DOUBLE",
+      "label_en": "Bodies founded in clandestine graves",
+      "label_es": "Cuerpos / restos recuperados",
+      "unit": "count",
+      "source": "Plataforma Ciudadana de Fosas, 2006-2024",
+      "range": [0, 154],
+      "aliases": ["bodies", "remains", "cuerpos", "restos"]
+    },
+    {
+      "key": "pam_n",
+      "table": "dim_pam_mun",
+      "type": "INTEGER",
+      "label_en": "Irregular migration events (PAM)",
+      "label_es": "Eventos de migracion irregular (PAM)",
+      "unit": "count",
+      "source": "Unidad de Politica Migratoria (PAM), Gobierno de Mexico",
+      "range": [1, 17464],
+      "aliases": ["irregular migration", "pam", "migracion irregular", "presentados"]
+    },
+    {
+      "key": "pam_presentados_n",
+      "table": "dim_pam_mun",
+      "type": "INTEGER",
+      "label_en": "Irregular migration cases",
+      "label_es": "PAM personas presentadas",
+      "unit": "count",
+      "source": "Unidad de Politica Migratoria (PAM), Gobierno de Mexico",
+      "range": [0, 17357],
+      "aliases": "presentados"
+    },
+    {
+      "key": "pam_canalizados_n",
+      "table": "dim_pam_mun",
+      "type": "INTEGER",
+      "label_en": "Irregular migration deported",
+      "label_es": "PAM personas canalizadas",
+      "unit": "count",
+      "source": "Unidad de Politica Migratoria (PAM), Gobierno de Mexico",
+      "range": [0, 5440],
+      "aliases": "canalizados"
+    },
+    {
+      "key": "pam_hombre_n",
+      "table": "dim_pam_mun",
+      "type": "INTEGER",
+      "label_en": "Irregular migrates cases (male)",
+      "label_es": "PAM hombres",
+      "unit": "count",
+      "source": "Unidad de Politica Migratoria (PAM), Gobierno de Mexico",
+      "range": [0, 13708],
+      "aliases": ["men", "hombres"]
+    },
+    {
+      "key": "pam_mujer_n",
+      "table": "dim_pam_mun",
+      "type": "INTEGER",
+      "label_en": "Irregular migrates cases (female)",
+      "label_es": "PAM mujeres",
+      "unit": "count",
+      "source": "Unidad de Politica Migratoria (PAM), Gobierno de Mexico",
+      "range": [0, 3967],
+      "aliases": ["women", "mujeres"]
+    },
+    {
+      "key": "pob_muj",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [41.4818, 59.4937],
+      "aliases": []
+    },
+    {
+      "key": "pob_hom",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [40.5063, 58.5182],
+      "aliases": []
+    },
+    {
+      "key": "men_5",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [2.596, 18.3688],
+      "aliases": []
+    },
+    {
+      "key": "edad_5_14",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [5.8824, 31.3135],
+      "aliases": []
+    },
+    {
+      "key": "edad_15_59",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [42.3464, 70.7873],
+      "aliases": []
+    },
+    {
+      "key": "edad_60_74",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [2.218, 26.9634],
+      "aliases": []
+    },
+    {
+      "key": "edad_75",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [0.587, 19.4656],
+      "aliases": []
+    },
+    {
+      "key": "pob_afro",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [0, 95.6911],
+      "aliases": []
+    },
+    {
+      "key": "decl_g",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [0, 4],
+      "aliases": []
+    },
+    {
+      "key": "decl_h",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [0, 11],
+      "aliases": []
+    },
+    {
+      "key": "prc_anp",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [0, 100],
+      "aliases": []
+    },
+    {
+      "key": "pr_p_1hmas_bbienestar_auto",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [0, 100],
+      "aliases": []
+    },
+    {
+      "key": "pr_p_1hmas_bbienestar_pie",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [0, 100],
+      "aliases": []
+    },
+    {
+      "key": "ind_conec",
+      "table": "dim_patp_mun",
+      "type": "DOUBLE",
+      "label_en": "Roads connectivity index",
+      "label_es": "",
+      "unit": "",
+      "source": "CONEVAL - PATP territorial analysis, 2020",
+      "range": [0.0112, 0.9313],
+      "aliases": []
+    },
+    {
+      "key": "pob_2020",
+      "table": "dim_pobreza_mun",
+      "type": "INTEGER",
+      "label_en": "poverty",
+      "label_es": "pobreza",
+      "unit": "",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [81, 1913345],
+      "aliases": []
+    },
+    {
+      "key": "pobreza_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "Share population in poverty",
+      "label_es": "Porcentaje de población en pobreza",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [5.45, 99.65],
+      "aliases": []
+    },
+    {
+      "key": "pobreza_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [55, 816934],
+      "aliases": []
+    },
+    {
+      "key": "pobreza_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.21, 3.93],
+      "aliases": []
+    },
+    {
+      "key": "pobreza_ext_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "Share population in extreme poverty",
+      "label_es": "Porcentaje de población en pobreza extrema",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 84.45],
+      "aliases": []
+    },
+    {
+      "key": "pobreza_ext_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 126672],
+      "aliases": []
+    },
+    {
+      "key": "pobreza_ext_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [3.03, 4.26],
+      "aliases": []
+    },
+    {
+      "key": "pobreza_mod_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [5.17, 85.04],
+      "aliases": []
+    },
+    {
+      "key": "pobreza_mod_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [47, 700991],
+      "aliases": []
+    },
+    {
+      "key": "pobreza_mod_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.2, 3.59],
+      "aliases": []
+    },
+    {
+      "key": "vul_carencia_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 77.58],
+      "aliases": []
+    },
+    {
+      "key": "vul_carencia_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 728225],
+      "aliases": []
+    },
+    {
+      "key": "vul_carencia_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.17, 3.39],
+      "aliases": []
+    },
+    {
+      "key": "vul_ingreso_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 23.61],
+      "aliases": []
+    },
+    {
+      "key": "vul_ingreso_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 215576],
+      "aliases": []
+    },
+    {
+      "key": "no_pobre_no_vul_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 57.43],
+      "aliases": []
+    },
+    {
+      "key": "no_pobre_no_vul_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 615131],
+      "aliases": []
+    },
+    {
+      "key": "rezago_edu_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [2.87, 61.39],
+      "aliases": []
+    },
+    {
+      "key": "rezago_edu_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [16, 319095],
+      "aliases": []
+    },
+    {
+      "key": "rezago_edu_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.36, 4.4],
+      "aliases": []
+    },
+    {
+      "key": "caren_salud_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.05, 83.86],
+      "aliases": []
+    },
+    {
+      "key": "caren_salud_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [12, 637666],
+      "aliases": []
+    },
+    {
+      "key": "caren_salud_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.93, 4.77],
+      "aliases": []
+    },
+    {
+      "key": "caren_seg_social_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [22.03, 96.99],
+      "aliases": []
+    },
+    {
+      "key": "caren_seg_social_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [52, 955642],
+      "aliases": []
+    },
+    {
+      "key": "caren_seg_social_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.23, 3.96],
+      "aliases": []
+    },
+    {
+      "key": "caren_vivienda_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0.78, 76.68],
+      "aliases": []
+    },
+    {
+      "key": "caren_vivienda_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [5, 130963],
+      "aliases": []
+    },
+    {
+      "key": "caren_vivienda_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.76, 4.62],
+      "aliases": []
+    },
+    {
+      "key": "caren_serv_basicos_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0.08, 100],
+      "aliases": []
+    },
+    {
+      "key": "caren_serv_basicos_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1, 220919],
+      "aliases": []
+    },
+    {
+      "key": "caren_serv_basicos_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.51, 4.38],
+      "aliases": []
+    },
+    {
+      "key": "caren_alimentacion_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 75.66],
+      "aliases": []
+    },
+    {
+      "key": "caren_alimentacion_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0, 515767],
+      "aliases": []
+    },
+    {
+      "key": "caren_alimentacion_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.73, 4.58],
+      "aliases": []
+    },
+    {
+      "key": "min_1_carencia_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [38.85, 100],
+      "aliases": []
+    },
+    {
+      "key": "min_1_carencia_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [75, 1220637],
+      "aliases": []
+    },
+    {
+      "key": "min_1_carencia_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.19, 3.92],
+      "aliases": []
+    },
+    {
+      "key": "min_3_carencias_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.2, 89.89],
+      "aliases": []
+    },
+    {
+      "key": "min_3_carencias_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [6, 303656],
+      "aliases": []
+    },
+    {
+      "key": "min_3_carencias_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [3.04, 4.23],
+      "aliases": []
+    },
+    {
+      "key": "ing_inf_lp_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [7.25, 99.89],
+      "aliases": []
+    },
+    {
+      "key": "ing_inf_lp_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [56, 1020408],
+      "aliases": []
+    },
+    {
+      "key": "ing_inf_lp_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0.72, 3.93],
+      "aliases": []
+    },
+    {
+      "key": "ing_inf_lpe_pct",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "%",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [1.09, 97.55],
+      "aliases": []
+    },
+    {
+      "key": "ing_inf_lpe_pers",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "people",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [11, 362871],
+      "aliases": []
+    },
+    {
+      "key": "ing_inf_lpe_carencias_prom",
+      "table": "dim_pobreza_mun",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "avg deprivations",
+      "source": "CONEVAL - Municipal poverty measurement, 2020",
+      "range": [0.71, 3.95],
+      "aliases": []
+    },
+    {
+      "key": "pobtot",
+      "table": "dim_pop_cpv_grid",
+      "type": "DOUBLE",
+      "label_en": "Population",
+      "label_es": "Población",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [0, 24585342],
+      "aliases": ["total population", "poblacion total", "pobtot"]
+    },
+    {
+      "key": "pnacop",
+      "table": "dim_pop_cpv_grid",
+      "type": "DOUBLE",
+      "label_en": "Population born abroad (Census 2020)",
+      "label_es": "Poblacion nacida en el extranjero",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 170217],
+      "aliases": ["born abroad", "foreign born", "nacida en el extranjero", "pnacop"]
+    },
+    {
+      "key": "presop2015",
+      "table": "dim_pop_cpv_grid",
+      "type": "DOUBLE",
+      "label_en": "Resided abroad in 2015 (Census 2020)",
+      "label_es": "Residian en otro pais en 2015",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 77184],
+      "aliases": ["resided abroad 2015", "presop2015"]
+    },
+    {
+      "key": "remesas_2024",
+      "table": "dim_remesas_mun",
+      "type": "DOUBLE",
+      "label_en": "Remittances 2024",
+      "label_es": "Remesas 2024",
+      "unit": "MXN",
+      "source": "CONAPO - migration intensity / remittances",
+      "range": [0, 908.6767],
+      "aliases": ["remittances", "remesas"]
+    },
+    {
+      "key": "year",
+      "table": "ged_events",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "UCDP Georeferenced Event Dataset (GED), 2019-2024",
+      "range": [2019, 2024],
+      "aliases": []
+    },
+    {
+      "key": "priogrid_gid",
+      "table": "ged_events",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "UCDP Georeferenced Event Dataset (GED), 2019-2024",
+      "range": [150656, 176531],
+      "aliases": []
+    },
+    {
+      "key": "best",
+      "table": "ged_events",
+      "type": "DOUBLE",
+      "label_en": "GED conflict deaths (best estimate)",
+      "label_es": "Muertes por conflicto GED (mejor estim.)",
+      "unit": "count",
+      "source": "UCDP Georeferenced Event Dataset (GED), 2019-2024",
+      "range": [0, 145],
+      "aliases": ["ged", "ucdp", "conflict deaths", "conflicto"]
+    },
+    {
+      "key": "TiempoViaje",
+      "table": "graves",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "Geospatial analysis of clandestine graves (Reinforcement Learning)",
+      "range": [0, 165.6634],
+      "aliases": []
+    },
+    {
+      "key": "Visibilidad",
+      "table": "graves",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "Geospatial analysis of clandestine graves (Reinforcement Learning)",
+      "range": [0.4, 56.62],
+      "aliases": []
+    },
+    {
+      "key": "pobtot",
+      "table": "grid_nivel4",
+      "type": "DOUBLE",
+      "label_en": "Population",
+      "label_es": "Población",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [0, 24585342],
+      "aliases": ["total population", "poblacion total", "pobtot"]
+    },
+    {
+      "key": "pnacop",
+      "table": "grid_nivel4",
+      "type": "DOUBLE",
+      "label_en": "Population born abroad (Census 2020)",
+      "label_es": "Poblacion nacida en el extranjero",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-6, 170217],
+      "aliases": ["born abroad", "foreign born", "nacida en el extranjero", "pnacop"]
+    },
+    {
+      "key": "presop2015",
+      "table": "grid_nivel4",
+      "type": "DOUBLE",
+      "label_en": "Resided abroad in 2015 (Census 2020)",
+      "label_es": "Residian en otro pais en 2015",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-6, 77184],
+      "aliases": ["resided abroad 2015", "presop2015"]
+    },
+    {
+      "key": "pobtot",
+      "table": "grid_nivel5",
+      "type": "DOUBLE",
+      "label_en": "Population",
+      "label_es": "Población",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [0, 9029039],
+      "aliases": ["total population", "poblacion total", "pobtot"]
+    },
+    {
+      "key": "pnacop",
+      "table": "grid_nivel5",
+      "type": "DOUBLE",
+      "label_en": "Population born abroad (Census 2020)",
+      "label_es": "Poblacion nacida en el extranjero",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 72380],
+      "aliases": ["born abroad", "foreign born", "nacida en el extranjero", "pnacop"]
+    },
+    {
+      "key": "presop2015",
+      "table": "grid_nivel5",
+      "type": "DOUBLE",
+      "label_en": "Resided abroad in 2015 (Census 2020)",
+      "label_es": "Residian en otro pais en 2015",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 32847],
+      "aliases": ["resided abroad 2015", "presop2015"]
+    },
+    {
+      "key": "pobtot",
+      "table": "grid_nivel6",
+      "type": "DOUBLE",
+      "label_en": "Population",
+      "label_es": "Población",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [0, 2622896],
+      "aliases": ["total population", "poblacion total", "pobtot"]
+    },
+    {
+      "key": "pnacop",
+      "table": "grid_nivel6",
+      "type": "DOUBLE",
+      "label_en": "Population born abroad (Census 2020)",
+      "label_es": "Poblacion nacida en el extranjero",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 44895],
+      "aliases": ["born abroad", "foreign born", "nacida en el extranjero", "pnacop"]
+    },
+    {
+      "key": "presop2015",
+      "table": "grid_nivel6",
+      "type": "DOUBLE",
+      "label_en": "Resided abroad in 2015 (Census 2020)",
+      "label_es": "Residian en otro pais en 2015",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 20696],
+      "aliases": ["resided abroad 2015", "presop2015"]
+    },
+    {
+      "key": "pobtot",
+      "table": "grid_nivel7",
+      "type": "DOUBLE",
+      "label_en": "Population",
+      "label_es": "Población",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [0, 367487],
+      "aliases": ["total population", "poblacion total", "pobtot"]
+    },
+    {
+      "key": "pnacop",
+      "table": "grid_nivel7",
+      "type": "DOUBLE",
+      "label_en": "Population born abroad (Census 2020)",
+      "label_es": "Poblacion nacida en el extranjero",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 17987],
+      "aliases": ["born abroad", "foreign born", "nacida en el extranjero", "pnacop"]
+    },
+    {
+      "key": "presop2015",
+      "table": "grid_nivel7",
+      "type": "DOUBLE",
+      "label_en": "Resided abroad in 2015 (Census 2020)",
+      "label_es": "Residian en otro pais en 2015",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 9024],
+      "aliases": ["resided abroad 2015", "presop2015"]
+    },
+    {
+      "key": "pobtot",
+      "table": "grid_nivel8",
+      "type": "DOUBLE",
+      "label_en": "Population",
+      "label_es": "Población",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [0, 65588],
+      "aliases": ["total population", "poblacion total", "pobtot"]
+    },
+    {
+      "key": "pnacop",
+      "table": "grid_nivel8",
+      "type": "DOUBLE",
+      "label_en": "Population born abroad (Census 2020)",
+      "label_es": "Poblacion nacida en el extranjero",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 4702],
+      "aliases": ["born abroad", "foreign born", "nacida en el extranjero", "pnacop"]
+    },
+    {
+      "key": "presop2015",
+      "table": "grid_nivel8",
+      "type": "DOUBLE",
+      "label_en": "Resided abroad in 2015 (Census 2020)",
+      "label_es": "Residian en otro pais en 2015",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI), 2020",
+      "range": [-7, 2526],
+      "aliases": ["resided abroad 2015", "presop2015"]
+    },
+    {
+      "key": "ALTITUD",
+      "table": "localities",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "Censo de Poblacion y Vivienda (INEGI) - localities, 2020",
+      "range": [-11, 3498],
+      "aliases": []
+    },
+    {
+      "key": "POB_TOTAL",
+      "table": "localities",
+      "type": "INTEGER",
+      "label_en": "Population",
+      "label_es": "Población",
+      "unit": "people",
+      "source": "Censo de Poblacion y Vivienda (INEGI) - localities, 2020",
+      "range": [1, 1835486],
+      "aliases": ["locality population", "poblacion localidad"]
+    },
+    {
+      "key": "PNACOE",
+      "table": "localities",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "Censo de Poblacion y Vivienda (INEGI) - localities, 2020",
+      "range": [0, 789327],
+      "aliases": []
+    },
+    {
+      "key": "PRES2015",
+      "table": "localities",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "Censo de Poblacion y Vivienda (INEGI) - localities, 2020",
+      "range": [1, 1673257],
+      "aliases": []
+    },
+    {
+      "key": "source_quality",
+      "table": "migrants",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "Missing Migrants Project (IOM), 2014-2025",
+      "range": [1, 5],
+      "aliases": []
+    },
+    {
+      "key": "n_missing",
+      "table": "migrants",
+      "type": "DOUBLE",
+      "label_en": "Migrants missing",
+      "label_es": "Migrantes desaparecidos",
+      "unit": "count",
+      "source": "Missing Migrants Project (IOM), 2014-2025",
+      "range": [-2, 60],
+      "aliases": ["missing migrants", "migrantes desaparecidos"]
+    },
+    {
+      "key": "n_dead",
+      "table": "migrants",
+      "type": "DOUBLE",
+      "label_en": "Migrant deaths",
+      "label_es": "Migrantes fallecidos",
+      "unit": "count",
+      "source": "Missing Migrants Project (IOM), 2014-2025",
+      "range": [1, 123],
+      "aliases": ["migrant deaths", "fallecidos", "muertes"]
+    },
+    {
+      "key": "n_total",
+      "table": "migrants",
+      "type": "DOUBLE",
+      "label_en": "Death during migration",
+      "label_es": "Migrantes muertos o desaparecidos (total)",
+      "unit": "count",
+      "source": "Missing Migrants Project (IOM), 2014-2025",
+      "range": [1, 123],
+      "aliases": ["dead or missing", "muertos o desaparecidos"]
+    },
+    {
+      "key": "n_children",
+      "table": "migrants",
+      "type": "DOUBLE",
+      "label_en": "Migrant child casualties",
+      "label_es": "Menores migrantes (victimas)",
+      "unit": "count",
+      "source": "Missing Migrants Project (IOM), 2014-2025",
+      "range": [1, 10],
+      "aliases": ["children", "menores"]
+    },
+    {
+      "key": "n_females",
+      "table": "migrants",
+      "type": "DOUBLE",
+      "label_en": "Migrant female casualties",
+      "label_es": "Mujeres migrantes (victimas)",
+      "unit": "count",
+      "source": "Missing Migrants Project (IOM), 2014-2025",
+      "range": [0, 15],
+      "aliases": ["females", "mujeres"]
+    },
+    {
+      "key": "n_males",
+      "table": "migrants",
+      "type": "DOUBLE",
+      "label_en": "Migrant male casualties",
+      "label_es": "Hombres migrantes (victimas)",
+      "unit": "count",
+      "source": "Missing Migrants Project (IOM), 2014-2025",
+      "range": [0, 50],
+      "aliases": ["males", "hombres"]
+    },
+    {
+      "key": "n_survivors",
+      "table": "migrants",
+      "type": "DOUBLE",
+      "label_en": "Migrant survivors",
+      "label_es": "Migrantes sobrevivientes",
+      "unit": "count",
+      "source": "Missing Migrants Project (IOM), 2014-2025",
+      "range": [0, 346],
+      "aliases": ["survivors", "sobrevivientes"]
+    },
+    {
+      "key": "incident_year",
+      "table": "migrants",
+      "type": "INTEGER",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "Missing Migrants Project (IOM), 2014-2025",
+      "range": [2014, 2026],
+      "aliases": []
+    },
+    {
+      "key": "year",
+      "table": "ocved_events",
+      "type": "INTEGER",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "OCVED (Organized Criminal Violence Event Data), 2010-2018",
+      "range": [2010, 2018],
+      "aliases": []
+    },
+    {
+      "key": "month",
+      "table": "ocved_events",
+      "type": "INTEGER",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "OCVED (Organized Criminal Violence Event Data), 2010-2018",
+      "range": [0, 12],
+      "aliases": []
+    },
+    {
+      "key": "day",
+      "table": "ocved_events",
+      "type": "INTEGER",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "OCVED (Organized Criminal Violence Event Data), 2010-2018",
+      "range": [0, 31],
+      "aliases": []
+    },
+    {
+      "key": "state",
+      "table": "ocved_events",
+      "type": "INTEGER",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "OCVED (Organized Criminal Violence Event Data), 2010-2018",
+      "range": [1, 32],
+      "aliases": []
+    },
+    {
+      "key": "mun",
+      "table": "ocved_events",
+      "type": "INTEGER",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "OCVED (Organized Criminal Violence Event Data), 2010-2018",
+      "range": [1001, 32058],
+      "aliases": []
+    },
+    {
+      "key": "counter",
+      "table": "ocved_events",
+      "type": "INTEGER",
+      "label_en": "Organised crime events",
+      "label_es": "Eventos de violencia OCVED",
+      "unit": "count",
+      "source": "OCVED (Organized Criminal Violence Event Data), 2010-2018",
+      "range": [1, 81],
+      "aliases": ["ocved", "organized crime violence", "violencia organizada"]
+    },
+    {
+      "key": "length_m",
+      "table": "roads",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "Road network (INEGI / maps)",
+      "range": [0.0167, 215977.7978],
+      "aliases": []
+    },
+    {
+      "key": "dark_figures_perc",
+      "table": "state_polygons",
+      "type": "DOUBLE",
+      "label_en": "Dark figure (unreported crime)",
+      "label_es": "Cifra negra (delitos no denunciados)",
+      "unit": "%",
+      "source": "ENVIPE, 2022-2023",
+      "range": [88.2671, 97.1171],
+      "aliases": ["dark figure", "unreported crime", "cifra negra"]
+    },
+    {
+      "key": "perception_insecurity_perc",
+      "table": "state_polygons",
+      "type": "DOUBLE",
+      "label_en": "Perceived insecurity",
+      "label_es": "Percepcion de inseguridad",
+      "unit": "%",
+      "source": "ENVIPE, 2022-2023",
+      "range": [20.419, 57.993],
+      "aliases": ["insecurity perception", "percepcion de inseguridad", "inseguridad"]
+    },
+    {
+      "key": "SHAPE_len",
+      "table": "viaferrea",
+      "type": "DOUBLE",
+      "label_en": "",
+      "label_es": "",
+      "unit": "",
+      "source": "Railway network (INEGI / maps)",
+      "range": [1.2041, 167105.2372],
+      "aliases": []
+    }
+  ]
+}