Skip to content

Fix Helm reranker endpoint configuration - #2480

Open
nvwizhu wants to merge 2 commits into
NVIDIA:mainfrom
nvwizhu:agent/fix-helm-reranker-config
Open

Fix Helm reranker endpoint configuration#2480
nvwizhu wants to merge 2 commits into
NVIDIA:mainfrom
nvwizhu:agent/fix-helm-reranker-config

Conversation

@nvwizhu

@nvwizhu nvwizhu commented Aug 13, 2026

Copy link
Copy Markdown

Description

The Retriever service supports nim_endpoints.rerank_invoke_url and nim_endpoints.rerank_model_name, but the Helm chart did not expose or render corresponding values. As a result, reranker settings supplied through Helm were not included in retriever-service.yaml, and /v1/query requests with rerank=true could return HTTP 400.

This change:

  • adds serviceConfig.nimEndpoints.rerankInvokeUrl and serviceConfig.nimEndpoints.rerankModelName
  • renders them as nim_endpoints.rerank_invoke_url and nim_endpoints.rerank_model_name
  • covers standalone and all split service ConfigMaps
  • fails Helm rendering when a model is configured without a reranker URL
  • adds structured Helm render regression coverage and documents the two values

Validation

  • helm lint --strict nemo_retriever/helm
  • pytest -q nemo_retriever/tests/test_helm_rerank_endpoint.py — 5 passed
  • full Helm and Helm harness test suite — 171 passed, 32 subtests passed
  • git diff --check

Checklist

  • I am familiar with the Contributing Guidelines.
  • New or existing tests cover these changes.
  • The documentation is up to date with these changes.

@copy-pr-bot

copy-pr-bot Bot commented Aug 13, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

Signed-off-by: wizhu <wizhu@nvidia.com>
@nvwizhu
nvwizhu force-pushed the agent/fix-helm-reranker-config branch from 7e7e878 to 7be4920 Compare August 13, 2026 17:32
@nvwizhu
nvwizhu marked this pull request as ready for review August 13, 2026 17:34
@nvwizhu
nvwizhu requested review from a team as code owners August 13, 2026 17:34
@nvwizhu
nvwizhu requested a review from nkmcalli August 13, 2026 17:34
@greptile-apps

greptile-apps Bot commented Aug 13, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

The PR exposes remote reranker endpoint and model settings through Helm, renders them into every service topology, and rejects model-only configurations before deployment.

  • Adds reranker URL and model defaults and documentation.
  • Renders both fields in standalone and split ConfigMaps.
  • Adds regression coverage for supported configurations and the invalid model-only case.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains; the prior model-only configuration now fails during Helm rendering for both missing and whitespace-only endpoints.

Important Files Changed

Filename Overview
nemo_retriever/helm/templates/configmap.yaml Renders reranker settings across all service ConfigMaps and fail-fast rejects the previously reported model-only configuration.
nemo_retriever/helm/values.yaml Adds empty defaults for the two new user-configurable reranker values.
nemo_retriever/helm/README.md Documents the endpoint/model relationship and explicitly states that the endpoint is not auto-resolved.
nemo_retriever/tests/test_helm_rerank_endpoint.py Covers standalone and split rendering, URL-only configuration, and absent or whitespace-only endpoints paired with a model.

Reviews (2): Last reviewed commit: "Validate Helm reranker model configurati..." | Re-trigger Greptile

Comment thread nemo_retriever/helm/templates/configmap.yaml
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant