diff --git a/inference-configs/README.md b/inference-configs/README.md
new file mode 100644
index 0000000..1152b4d
--- /dev/null
+++ b/inference-configs/README.md
@@ -0,0 +1,28 @@
+# Inference Configuration YAML files
+
+This directory is meant to store the YAML files which are used for running [the imagery pipeline](https://github.com/SWFSC/esdglider/blob/57f349ddb22808d5ec548e4e469758f9a92532fa/docs/imagery-pipeline.md).
+
+Copy [template.yaml](template.yaml) into the associated deployment year folder,
+creating it if needed, and update the deployment and model. For example, save a
+2026 deployment's configuration as `2026/glider-imagery.yaml`. Use the year in
+the deployment's date, not the date you run processing.
+
+Pass the saved file's path to the pipeline's `--config` option, for example
+`~/glider-processing/inference-configs/2026/glider-imagery.yaml`.
+Commit and push it normally; the pipeline does not automate GitHub uploads or
+commits. Leave the template unchanged for other users to copy.
+
+Each new live submission to the imagery pipeline automatically stores
+the exact YAML and resolved local configuration in GCS, along with snapshots
+of the selected manifests and pipeline YAML:
+
+```text
+gs://{output_bucket}/SWFSC/glider/{deployment_name}/runs/{run_id}/config/
+gs://{output_bucket}/SWFSC/glider/{deployment_name}/runs/{run_id}/manifests/
+gs://{output_bucket}/SWFSC/glider/{deployment_name}/runs/{run_id}/outputs/{chunk_number:02d}/
+```
+
+The run ID contains a UTC timestamp and UUID. Archive writes are create-only
+at the application level; the bucket does not enforce immutability. Use the
+printed run ID with `--resume-run-id` to resume DAG submission without rewriting
+the archive or regenerating metadata.
diff --git a/inference-configs/template.yaml b/inference-configs/template.yaml
new file mode 100644
index 0000000..fc59f1e
--- /dev/null
+++ b/inference-configs/template.yaml
@@ -0,0 +1,7 @@
+# Copy into the deployment's year folder, then replace the deployment and model.
+# deployment_name must use the real glider name and an eight-digit YYYYMMDD date.
+deployment_name: glider-YYYYMMDD
+
+dag:
+  # Replace with the approved model ID for your data.
+  model_type: CHOOSE_MODEL
diff --git a/inference-configs/test_run_records.yaml b/inference-configs/test_run_records.yaml
new file mode 100644
index 0000000..e8f5e9f
--- /dev/null
+++ b/inference-configs/test_run_records.yaml
@@ -0,0 +1,4 @@
+# Test config to see whether the run records work with GCS DAG permissions
+deployment_name: amlr04-20230620
+dag:
+  model_type: andy-6d-v1-detector-model