mcp-server/index.ts

@@ -151,7 +151,9 @@ server.tool(
   },
   async ({ question, sql_filter, dataset, id_column, top_k }) => {
     const body: any = {
-      question, index_name: "workers_500k_v1", filter_dataset: dataset, id_column, top_k, generate: true,
+      // Index must share an id-space with filter_dataset (default ethereal_workers, 10K)
+      // → ethereal_workers_v1. workers_500k_v1 (50K, different workers) misaligned.
+      question, index_name: "ethereal_workers_v1", filter_dataset: dataset, id_column, top_k, generate: true,
       use_playbook_memory: true,
     };
     if (sql_filter) body.sql_filter = sql_filter;
@@ -185,7 +187,8 @@ server.tool(
     if (city) filter += ` AND city = '${city}'`;
     const r = await api("POST", "/vectors/hybrid", {
       question: `Find the best ${role} workers with relevant skills and certifications`,
-      index_name: "workers_500k_v1", sql_filter: filter,
+      // ethereal_workers_v1 aligns with filter_dataset ethereal_workers (same 10K ids).
+      index_name: "ethereal_workers_v1", sql_filter: filter,
       filter_dataset: "ethereal_workers", id_column: "worker_id",
       top_k: headcount * 2, generate: false,
       use_playbook_memory: true,
@@ -556,7 +559,8 @@ async function main() {
           if (b.city) filter += ` AND city = '${b.city}'`;
           return ok(await api("POST", "/vectors/hybrid", {
             question: `Best ${b.role} workers with relevant skills`,
-            index_name: b.index || "workers_500k_v1", sql_filter: filter,
+            // Align default index with the ethereal_workers dataset (see /search fix).
+            index_name: b.index || "ethereal_workers_v1", sql_filter: filter,
             filter_dataset: b.dataset || "ethereal_workers",
             id_column: "worker_id", top_k: (b.headcount || 5) * 2, generate: false,
             use_playbook_memory: true,