fix(search): scope index create to managed databases; guard sorted in run

Close the create/show/remove asymmetry the review flagged: `search create`
resolved any catalog `connections::resolve_connection_id` matched — including a
plain connection — while `search show`/`search remove` (via `locate_by_name`)
require a managed database. An index created on a connection was therefore
listable but neither showable nor removable, and top-level `indexes delete` is
gone in this PR, so it had no deletion path.

Indexes are a managed-database concept (connections are a legacy concept being
removed), so `create` now resolves `--from`'s catalog through
`databases::try_resolve_database` and uses the database's own
`default_connection_id`; a plain connection is rejected with a clear message.
create/show/remove now agree on managed databases.

Also guard `search run` against a `sorted` index (not a retrieval type) — it
previously fell into the vector branch and emitted an opaque `vector_distance`
error; it now points the user at `hotdata query`.

Docs: search SKILL notes indexes are managed-database only and adds the
`search show` command block.

Author: eddietejeda

skills/hotdata/subskills/search/SKILL.md

@@ -47,20 +47,23 @@ hotdata search "<query>" --in <name>
 
 ## Indexes (text and vector)
 
-Create names the index (positional) and attaches to a table via `--from <catalog.schema.table>` (a managed database or an attached catalog). `list` narrows to the **active database** when one is set; without one it scans the whole workspace. `remove` takes the index name.
+Indexes are a **managed-database** concept. Create names the index (positional) and attaches to a table on a managed database via `--from` — `catalog.schema.table` (the managed database's catalog), or `schema.table` with an active database set. A plain connection catalog is rejected. `list` narrows to the **active database** when one is set; without one it scans the whole workspace. `show`/`remove` resolve the index by name in the active database (or `--database <id>`).
 
 ```bash
 # List — active-database scope when a DB is set, else whole-workspace scan
 hotdata search list [--workspace-id <ws>] [--output table|json|yaml]
 
-# Create — index name is positional; --from is catalog.schema.table
+# Create — index name is positional; --from is a managed database's table
 hotdata search create <name> --type text|vector --from <catalog.schema.table> \
   --column <col> \
   [--metric l2|cosine|dot] [--async] \
   [--provider <id>] [--dimensions <n>] [--output-column <name>] [--description <text>]
 
-# Remove — by index name
-hotdata search remove <name>
+# Show — by index name, in the active database (or -d/--database <id>)
+hotdata search show <name> [-d <db-id>] [--output table|json|yaml]
+
+# Remove — by index name, in the active database (or -d/--database <id>)
+hotdata search remove <name> [-d <db-id>]
 ```
 
 - **`--type` is required** on create: `text` (BM25; one or more text columns, comma-separated in `--column`) or `vector` (exactly one column; often embeddings or auto-embedded text). (`sorted` is also a valid `--type`, covered in **`hotdata-analytics`** — [`../analytics/SKILL.md`](../analytics/SKILL.md).)

src/commands/search.rs

@@ -9,7 +9,7 @@
 use crate::client::sdk::Api;
 use crate::commands::embedding_providers::{self, EmbeddingProvidersCommands};
 use crate::commands::indexes::{self, IndexScope};
-use crate::commands::{connections, databases, query};
+use crate::commands::{databases, query};
 
 /// Subcommands for `hotdata search`.
 #[derive(clap::Subcommand)]
@@ -240,9 +240,26 @@ fn create(
         "sorted" => "sorted",
         _ => "vector",
     };
-    let (conn_name, schema, table) = parse_table(workspace_id, from);
+    let (catalog, schema, table) = parse_table(workspace_id, from);
     let api = Api::new(Some(workspace_id));
-    let conn_id = connections::resolve_connection_id(&api, &conn_name);
+    // Indexes are a managed-database concept. Resolve --from's catalog to a
+    // managed database and use its connection; reject a plain connection (a
+    // legacy concept being removed) so create can't make an index that the
+    // managed-database-only `search show`/`search remove` can't reach.
+    let db = databases::try_resolve_database(&api, &catalog).unwrap_or_else(|_| {
+        use crossterm::style::Stylize;
+        eprintln!(
+            "{}",
+            format!(
+                "error: '{catalog}' is not a managed database. Search indexes are created on \
+                 managed databases — pass a managed database's catalog, or 'schema.table' with an \
+                 active database set via 'hotdata databases use <id>'."
+            )
+            .red()
+        );
+        std::process::exit(1);
+    });
+    let conn_id = db.default_connection_id;
     let auto_name = format!("{table}_{}_{index_type}", column.replace(',', "_"));
     let index_name = name.unwrap_or(auto_name.as_str());
     indexes::create(
@@ -333,6 +350,18 @@ pub fn run(
     output: &str,
 ) {
     let loc = locate_or_exit(workspace_id, database, name);
+    if loc.index_type == "sorted" {
+        use crossterm::style::Stylize;
+        eprintln!(
+            "{}",
+            format!(
+                "error: index '{name}' is a sorted index — not searchable. Use 'hotdata query' \
+                 with a WHERE/ORDER BY filter instead."
+            )
+            .red()
+        );
+        std::process::exit(1);
+    }
     let table_fqn = format!("{}.{}.{}", loc.catalog, loc.schema, loc.table);
     let sql = build_search_sql(
         &loc.index_type,