diff --git a/stacks/fire-planner/main.tf b/stacks/fire-planner/main.tf
index 5c6b934a..6af26754 100644
--- a/stacks/fire-planner/main.tf
+++ b/stacks/fire-planner/main.tf
@@ -806,6 +806,10 @@ resource "kubernetes_job_v1" "examples_bulk_ingest" {
             name  = "LLM_MODEL"
             value = var.examples_llm_model
           }
+          env {
+            name  = "LLM_CONCURRENCY"
+            value = "3"
+          }
         }
       }
     }
diff --git a/stacks/fire-planner/terragrunt.hcl b/stacks/fire-planner/terragrunt.hcl
index c1d2e468..15fa4d4f 100644
--- a/stacks/fire-planner/terragrunt.hcl
+++ b/stacks/fire-planner/terragrunt.hcl
@@ -25,4 +25,9 @@ dependency "dbaas" {
 inputs = {
   # fire-planner repo HEAD — bump on every deploy.
   image_tag = "latest"
+
+  # Bulk ingest toggle — flip to true once, apply, monitor job, then reset to false.
+  run_examples_bulk_ingest = true
+  # qwen3-8b: GPU has ~10.7 GB free (immich-ml using ~4 GB of 15 GB total).
+  examples_llm_model = "qwen3-8b"
 }