Skip to content

change all references from 26.05 to 26.08 - #2574

Open
jperez999 wants to merge 1 commit into
NVIDIA:mainfrom
jperez999:release-26.08
Open

change all references from 26.05 to 26.08#2574
jperez999 wants to merge 1 commit into
NVIDIA:mainfrom
jperez999:release-26.08

Conversation

@jperez999

Copy link
Copy Markdown
Collaborator

Description

Updates references to 26.08 from 26.05.

Checklist

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

@jperez999 jperez999 self-assigned this Aug 24, 2026
@jperez999
jperez999 requested review from a team as code owners August 24, 2026 16:10
@jperez999
jperez999 requested a review from jdye64 August 24, 2026 16:10
@copy-pr-bot

copy-pr-bot Bot commented Aug 24, 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.

@greptile-apps

greptile-apps Bot commented Aug 24, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

This PR advances release, Helm, documentation, workflow, and compatibility references from 26.05 to 26.08 and adds an NVIDIA-hosted-endpoints service profile.

  • Updates the chart, service image, documentation site metadata, examples, and tests to 26.8.0.
  • Retargets manual documentation publishing to the 26.08 branch while retaining dry-run and explicit latest-alias controls.
  • Adds a service configuration for NVIDIA-hosted extraction, embedding, and reranking endpoints.

Confidence Score: 1/5

The PR is not safe to merge until the exposed MCP write path is secured, the new profile is packaged, and the ingest compatibility boundary is restored to its historically correct release.

The added profile permits unauthenticated network clients to initiate ingestion, is absent from installed wheel artifacts, and the changed runtime messages direct operators using the established 26.05 API boundary toward an incorrect 26.08 requirement.

Files Needing Attention: nemo_retriever/src/nemo_retriever/service/retriever-service-nvidia-endpoints.yaml, nemo_retriever/src/nemo_retriever/service/client.py, nemo_retriever/src/nemo_retriever/service/routers/ingest.py, nemo_retriever/pyproject.toml

Security Review

The new NVIDIA-endpoints profile binds publicly, disables bearer authentication, and enables MCP ingestion tools, allowing unauthenticated state-changing requests whenever an operator launches this profile on a reachable host.

Important Files Changed

Filename Overview
.github/workflows/nrl-docs-nvidia-publish.yml Retargets the deliberate, dry-run-first publishing workflow and latest metadata to 26.08 without restoring unsafe automatic triggers.
docs/publish/versions.json Moves the latest documentation alias from 26.5.0 to 26.8.0 consistently with the workflow source branch.
nemo_retriever/src/nemo_retriever/service/client.py Incorrectly relabels the historical job-scoped ingest API compatibility boundary as 26.08.
nemo_retriever/src/nemo_retriever/service/routers/ingest.py Incorrectly reports that legacy ingest routes were removed in 26.08 rather than 26.05.
nemo_retriever/src/nemo_retriever/service/retriever-service-nvidia-endpoints.yaml Adds an unpackaged profile that exposes unauthenticated MCP write operations when explicitly launched on a reachable host.
nemo_retriever/helm/values.yaml Updates the default service image and release-specific comments to 26.8.0.
Prompt To Fix All With AI
### Issue 1
nemo_retriever/src/nemo_retriever/service/client.py:171-173
**Incorrect ingest compatibility boundary**

When an SDK/service mismatch returns 404 or 410, this message says the job-scoped API was introduced in 26.08 and requires a 26.08+ service, but the API and legacy-route stubs entered through the 26.05 merge. This obscures the actual compatibility boundary and can prompt an unnecessary upgrade.

### Issue 2
nemo_retriever/src/nemo_retriever/service/retriever-service-nvidia-endpoints.yaml:219-225
**Unauthenticated MCP write operations**

If an operator launches this profile on a network-reachable host, it binds to `0.0.0.0` without bearer authentication while registering the `ingest_documents` MCP tool, allowing unauthenticated clients to create jobs, upload documents, and consume service or hosted-inference resources.

**How this was verified:** The profile's listener and auth settings were traced through MCP registration to the job-creation and document-upload endpoints.

### Issue 3
nemo_retriever/src/nemo_retriever/service/retriever-service-nvidia-endpoints.yaml:1
**Service profile omitted from wheel**

When a user installs the published package and selects this NVIDIA-endpoints profile, the file is unavailable because the service package-data manifest includes only `retriever-service.yaml`. Starting the service with the new profile path therefore fails unless the user separately obtains the source file.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Reviews (1): Last reviewed commit: "change all references from 26.05 to 26.0..." | Re-trigger Greptile

Comment on lines +171 to +173
"+ GET /v1/ingest/job/{job_id}/events) introduced in 26.08, "
"but the deployed nrl-service does not advertise that route. "
"Upgrade the chart/image to a 26.05+ build, or downgrade the "
"Upgrade the chart/image to a 26.08+ build, or downgrade the "

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Incorrect ingest compatibility boundary

When an SDK/service mismatch returns 404 or 410, this message says the job-scoped API was introduced in 26.08 and requires a 26.08+ service, but the API and legacy-route stubs entered through the 26.05 merge. This obscures the actual compatibility boundary and can prompt an unnecessary upgrade.

Rule Used: When this PR changes user-facing code, configurati... (source)

Prompt To Fix With AI
This is a comment left during a code review.
Path: nemo_retriever/src/nemo_retriever/service/client.py
Line: 171-173

Comment:
**Incorrect ingest compatibility boundary**

When an SDK/service mismatch returns 404 or 410, this message says the job-scoped API was introduced in 26.08 and requires a 26.08+ service, but the API and legacy-route stubs entered through the 26.05 merge. This obscures the actual compatibility boundary and can prompt an unnecessary upgrade.

**Rule Used:** When this PR changes user-facing code, configurati... ([source](https://github.com/nvidia/nemo-retriever/blob/f2eed28a44b07956a9f1770ec5a469a9381baf27/.greptile/config.json))

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

Comment on lines +219 to +225
poll_interval_s: 2.0

# Controls whether (and how widely) client-supplied PipelineSpec
# overrides are honored on a per-request basis.
#
# reject — every client override is rejected; YAML-only config.
# allow_list — only audited "shape" keys may be overridden

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 security Unauthenticated MCP write operations

If an operator launches this profile on a network-reachable host, it binds to 0.0.0.0 without bearer authentication while registering the ingest_documents MCP tool, allowing unauthenticated clients to create jobs, upload documents, and consume service or hosted-inference resources.

How this was verified: The profile's listener and auth settings were traced through MCP registration to the job-creation and document-upload endpoints.

Prompt To Fix With AI
This is a comment left during a code review.
Path: nemo_retriever/src/nemo_retriever/service/retriever-service-nvidia-endpoints.yaml
Line: 219-225

Comment:
**Unauthenticated MCP write operations**

If an operator launches this profile on a network-reachable host, it binds to `0.0.0.0` without bearer authentication while registering the `ingest_documents` MCP tool, allowing unauthenticated clients to create jobs, upload documents, and consume service or hosted-inference resources.

**How this was verified:** The profile's listener and auth settings were traced through MCP registration to the job-creation and document-upload endpoints.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

@@ -0,0 +1,254 @@
# Retriever Service Mode — NVIDIA Inference API MCP profile

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Service profile omitted from wheel

When a user installs the published package and selects this NVIDIA-endpoints profile, the file is unavailable because the service package-data manifest includes only retriever-service.yaml. Starting the service with the new profile path therefore fails unless the user separately obtains the source file.

Prompt To Fix With AI
This is a comment left during a code review.
Path: nemo_retriever/src/nemo_retriever/service/retriever-service-nvidia-endpoints.yaml
Line: 1

Comment:
**Service profile omitted from wheel**

When a user installs the published package and selects this NVIDIA-endpoints profile, the file is unavailable because the service package-data manifest includes only `retriever-service.yaml`. Starting the service with the new profile path therefore fails unless the user separately obtains the source file.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

@kheiss-uwzoo
kheiss-uwzoo self-requested a review August 24, 2026 16:16
@jperez999 jperez999 mentioned this pull request Aug 25, 2026
3 tasks
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.

4 participants