Use measured query-only context across desktop CLI and MCP

crates/searchi-cli/src/main.rs

@@ -203,7 +203,7 @@ fn main() -> Result<()> {
             });
             let results = if let (Some(model), Some(runtime)) = (model_dir, embedding_worker) {
                 let mut service = trace.timed("cli.session_initialize", None, |_, _| {
-                    HybridSearch::open(&db, model, runtime, usize::from(cpu_threads))
+                    HybridSearch::open_query(&db, model, runtime, usize::from(cpu_threads))
                 })?;
                 service
                     .search_traced(&query, usize::from(limit), &options, &mut trace, None)

crates/searchi-cli/src/mcp.rs

@@ -450,7 +450,7 @@ pub fn serve(
                 "No semantic generation; run explicit indexing/embedding outside MCP first"
             );
 
-            let mut hybrid = HybridSearch::open(&db, models, worker, threads)?;
+            let mut hybrid = HybridSearch::open_query(&db, models, worker, threads)?;
 
             hybrid.warmup()?;