Fix Helm reranker endpoint configuration

nemo_retriever/helm/README.md

@@ -325,6 +325,8 @@ The retriever service picks up the in-cluster ASR endpoint when `nimOperator.aud
 | `serviceConfig.pipeline.batchWorkers`             | `48`    | Per-pod batch worker count. Refer to [Timeouts and alleviating ingest failures](#timeouts-and-alleviating-ingest-failures) if embed or pool errors appear under load. |
 | `serviceConfig.resources.maxUploadBytes`          | `500000000` | Maximum upload file size in bytes; requests exceeding the limit are rejected before buffering. |
 | `serviceConfig.nimEndpoints.*InvokeUrl`           | `""`    | Override the auto-resolved NIM Operator URL. Available knobs: `pageElementsInvokeUrl`, `tableStructureInvokeUrl`, `ocrInvokeUrl`, `embedInvokeUrl`, and `captionInvokeUrl` (refer to [Image captioning (Omni 30B)](#image-captioning-omni-30b)). |
+| `serviceConfig.nimEndpoints.rerankInvokeUrl`      | `""`    | `/v1/ranking` endpoint used by `POST /v1/query` when `rerank=true`. This value is not auto-resolved from `nimOperator.rerankqa`. |
+| `serviceConfig.nimEndpoints.rerankModelName`      | `""`    | Model id sent to `rerankInvokeUrl`; setting a model requires a non-empty reranker URL. |
 | `serviceConfig.nimEndpoints.captionModelName`     | `""`    | Model id sent to the remote VLM. Auto-set to `nvidia/nemotron-3-nano-omni-30b-a3b-reasoning` whenever a caption URL is resolved. |
 | `serviceConfig.llm.enabled`                         | `false` | Enables `POST /v1/answer`. Auto-flips to true when `nimOperator.answer_llm` is enabled and the operator URL resolves. |
 | `serviceConfig.llm.apiBase`                         | `""`    | OpenAI-compatible LLM base URL. Explicit value wins; otherwise `answer_llm` opt-in resolves to `http://answer-llm:8000/v1` by default. |

nemo_retriever/helm/templates/configmap.yaml

@@ -20,6 +20,11 @@ inherits the NIMService resource name, so the mapping is fixed:
 {{- $auth := $ctx.Values.serviceConfig.auth -}}
 {{- $scopeTokenSecret := $auth.scopeTokenSecret -}}
 {{- $internalAuth := $ctx.Values.serviceConfig.vectordb.internalAuth -}}
+{{- $rerankInvokeURL := $ctx.Values.serviceConfig.nimEndpoints.rerankInvokeUrl | default "" | trim -}}
+{{- $rerankModelName := $ctx.Values.serviceConfig.nimEndpoints.rerankModelName | default "" | trim -}}
+{{- if and $rerankModelName (not $rerankInvokeURL) -}}
+{{- fail "serviceConfig.nimEndpoints.rerankModelName requires serviceConfig.nimEndpoints.rerankInvokeUrl." -}}
+{{- end -}}
 {{- if and $auth.enabled (not $auth.apiToken) (not $scopeTokenSecret.name) -}}
 {{- fail "serviceConfig.auth.enabled=true requires serviceConfig.auth.scopeTokenSecret.name or serviceConfig.auth.apiToken." -}}
 {{- end -}}
@@ -109,6 +114,8 @@ nim_endpoints:
 {{- else }}
   embed_model_provider_prefix: null
 {{- end }}
+  rerank_invoke_url: {{ if .Values.serviceConfig.nimEndpoints.rerankInvokeUrl }}{{ .Values.serviceConfig.nimEndpoints.rerankInvokeUrl | quote }}{{ else }}null{{ end }}
+  rerank_model_name: {{ if .Values.serviceConfig.nimEndpoints.rerankModelName }}{{ .Values.serviceConfig.nimEndpoints.rerankModelName | quote }}{{ else }}null{{ end }}
   caption_invoke_url: {{ if .captionURL }}{{ .captionURL | quote }}{{ else }}null{{ end }}
   caption_model_name: {{ if .captionModelName }}{{ .captionModelName | quote }}{{ else }}null{{ end }}
   audio_grpc_endpoint: {{ if .audioGrpcEndpoint }}{{ .audioGrpcEndpoint | quote }}{{ else }}null{{ end }}

nemo_retriever/helm/values.yaml

@@ -552,6 +552,8 @@ serviceConfig:
     # nvidia/nemotron-parse or self-hosted v1.2 contract from the URL.
     nemotronParseModel: ""
     embedInvokeUrl: ""
+    rerankInvokeUrl: ""
+    rerankModelName: ""
     # Optional remote VLM endpoint for image captioning (Nemotron 3 Nano
     # Omni). Auto-wired from the in-cluster Service when
     # `nimOperator.nemotron_3_nano_omni_30b_a3b_reasoning.enabled=true`

nemo_retriever/tests/test_helm_rerank_endpoint.py

@@ -0,0 +1,72 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES.
+# All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+"""Helm wiring for remote reranker configuration."""
+
+import subprocess
+
+import pytest
+import yaml
+
+from tests.test_helm_shared_results import _render
+
+
+@pytest.mark.parametrize(
+    ("topology_args", "expected_modes"),
+    (((), {"standalone"}), (("--set", "topology.mode=split"), {"gateway", "realtime", "batch"})),
+)
+def test_rerank_endpoint_is_rendered_in_each_service_config(topology_args, expected_modes) -> None:
+    url = "http://reranker.example:8000/v1/ranking"
+    model = "nvidia/llama-nemotron-rerank-vl-1b-v2"
+    documents = _render(
+        *topology_args,
+        "--set-string",
+        f"serviceConfig.nimEndpoints.rerankInvokeUrl={url}",
+        "--set-string",
+        f"serviceConfig.nimEndpoints.rerankModelName={model}",
+    )
+
+    configs = [
+        yaml.safe_load(document["data"]["retriever-service.yaml"])
+        for document in documents
+        if document.get("kind") == "ConfigMap" and "retriever-service.yaml" in document.get("data", {})
+    ]
+
+    assert {config["mode"] for config in configs} == expected_modes
+    for config in configs:
+        assert config["nim_endpoints"]["rerank_invoke_url"] == url
+        assert config["nim_endpoints"]["rerank_model_name"] == model
+
+
+def test_rerank_endpoint_allows_url_without_model() -> None:
+    url = "http://reranker.example:8000/v1/ranking"
+    documents = _render(
+        "--set-string",
+        f"serviceConfig.nimEndpoints.rerankInvokeUrl={url}",
+    )
+
+    config = next(
+        yaml.safe_load(document["data"]["retriever-service.yaml"])
+        for document in documents
+        if document.get("kind") == "ConfigMap" and "retriever-service.yaml" in document.get("data", {})
+    )
+
+    assert config["nim_endpoints"]["rerank_invoke_url"] == url
+    assert config["nim_endpoints"]["rerank_model_name"] is None
+
+
+@pytest.mark.parametrize("endpoint", (None, "   "))
+def test_rerank_model_without_endpoint_fails_rendering(endpoint: str | None) -> None:
+    args = [] if endpoint is None else ["--set-string", f"serviceConfig.nimEndpoints.rerankInvokeUrl={endpoint}"]
+    with pytest.raises(subprocess.CalledProcessError) as error:
+        _render(
+            *args,
+            "--set-string",
+            "serviceConfig.nimEndpoints.rerankModelName=nvidia/llama-nemotron-rerank-vl-1b-v2",
+        )
+
+    assert (
+        "serviceConfig.nimEndpoints.rerankModelName requires serviceConfig.nimEndpoints.rerankInvokeUrl"
+        in error.value.stderr
+    )