Skip to content

fix(nvidia): use Nemotron Super 1M context - #98726

Merged
eleqtrizit merged 5 commits into
openclaw:mainfrom
eleqtrizit:765
Jul 1, 2026
Merged

fix(nvidia): use Nemotron Super 1M context#98726
eleqtrizit merged 5 commits into
openclaw:mainfrom
eleqtrizit:765

Conversation

@eleqtrizit

@eleqtrizit eleqtrizit commented Jul 1, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Summary

Correct the bundled NVIDIA Nemotron 3 Super metadata so OpenClaw's fallback catalog uses a 1,048,576-token context window.

Changes

  • Update the bundled Nemotron 3 Super context window from 262,144 to 1,048,576 tokens.
  • Leave NVIDIA's public featured-model feed values unchanged and authoritative.
  • Align the bundled fallback documentation and focused provider test.

Validation

  • node scripts/run-vitest.mjs extensions/nvidia passed.
  • Targeted oxfmt and oxlint checks passed.
  • NVIDIA docs MDX and link checks passed.
  • git diff --check passed.
  • Fresh tracked-diff autoreview reported no accepted/actionable findings (0.99 confidence).

Notes

  • NVIDIA can correct the public featured-model feed at its source; authenticated live-catalog consumers receive that value without OpenClaw normalization.
  • Nemotron 3 Ultra remains at 1,000,000 tokens.
  • Existing explicit model overrides and agents.defaults.contextTokens caps remain authoritative.
  • No config schema, CLI, gateway protocol, or Plugin SDK contract changes.
  • AI-assisted; I reviewed the implementation and validation results.

What Problem This Solves

The bundled fallback catalog advertised 262,144 tokens for Nemotron 3 Super even though the model supports a 1M-token context. Users relying on the bundled fallback could therefore compact earlier than necessary.

Why This Change Was Made

The bundled metadata now represents the model's 1M context as 1,048,576 tokens. The separate public featured-model feed remains provider-owned and passes through unchanged.

User Impact

Users of the bundled NVIDIA fallback catalog receive a 1,048,576-token context budget. Users receiving the live featured catalog continue to use NVIDIA's published feed value.

Evidence

  • NVIDIA documents Nemotron 3 Super with up to 1M tokens of context.
  • A focused provider test proves the bundled fallback exposes a 1,048,576-token context window.
  • Existing live-catalog tests pass with the provider-owned feed path unchanged.

@openclaw-barnacle openclaw-barnacle Bot added docs Improvements or additions to documentation extensions: nvidia size: S maintainer Maintainer-authored PR labels Jul 1, 2026
@clawsweeper

clawsweeper Bot commented Jul 1, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

Codex review: needs real behavior proof before merge. Reviewed July 1, 2026, 2:39 PM ET / 18:39 UTC.

Summary
The branch changes the NVIDIA Nemotron 3 Super bundled fallback catalog, docs table, and provider-catalog test from 262,144 to 1,048,576 context tokens.

PR surface: Source 0, Tests +4, Docs 0. Total +4 across 3 files.

Reproducibility: yes. for the source-level merge risk: current main and the live NVIDIA feed use 262,144 for Super, while PR head changes only the bundled fallback to 1,048,576. There is no high-confidence user-flow proof that the hosted OpenClaw NVIDIA path should use the larger fallback limit today.

Review metrics: 2 noteworthy metrics.

  • Fallback Context Metadata: 1 changed: 262,144 -> 1,048,576. This user-visible provider metadata feeds model picker and context budgeting when the live catalog is unavailable.
  • Hosted/Fallback Divergence: 1 model differs: hosted feed 262,144, fallback 1,048,576. Green CI does not settle whether fallback metadata may intentionally exceed NVIDIA's current hosted feed value.

Root-cause cluster
Relationship: same_root_cause
Canonical: #84228
Summary: This PR revisits the same Nemotron 3 Super fallback-context dispute as the earlier closed metadata PR, now with the final value set to 1,048,576 while keeping live feed values authoritative.

Members:

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🧂 unranked krab
Proof: 🧂 unranked krab
Patch quality: 🧂 unranked krab
Result: blocked until real behavior proof is added.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

  • Restore the Super fallback context to 262,144 unless maintainers explicitly sponsor the fallback-only long-context policy.
  • [P2] If maintainers sponsor the divergence, add redacted current-head terminal output, logs, live output, or a linked artifact showing the patched fallback catalog behavior.

Proof guidance:

  • [P1] Needs real behavior proof before merge: The PR body lists validation commands but no current-head terminal output, logs, live output, or linked artifact showing the patched fallback behavior; posted proof should be redacted and then the PR body can be updated for re-review. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Risk before merge

  • [P2] Merging would make the offline fallback row advertise 1,048,576 tokens while the current NVIDIA hosted feed still reports 262,144, so context-budget and compaction behavior can depend on feed availability.
  • [P2] This reverses the earlier closed same-change decision unless maintainers explicitly sponsor fallback-only long-context metadata as a product policy.
  • [P2] The PR body lists validation commands but does not include current-head terminal output, logs, live output, or a linked artifact showing the patched fallback behavior.

Maintainer options:

  1. Restore Hosted-Feed Alignment (recommended)
    Keep Nemotron 3 Super at 262,144 in the bundled fallback until NVIDIA's hosted feed or a maintainer-approved policy changes the contract.
  2. Sponsor Fallback-Only Long Context
    Maintainers can intentionally accept 1,048,576 for fallback-only support, but the PR should state that policy and include current-head runtime proof.
  3. Pause Or Close If Prior Decision Stands
    If the June hosted-feed-authority decision remains current, this PR should not stay open as a merge candidate.

Next step before merge

  • [P1] Manual review is needed because the protected maintainer label and prior hosted-feed decision make this a product/compatibility choice rather than an automation repair.

Security
Cleared: The diff changes static provider metadata, docs, and a focused test only; it adds no dependency, workflow, secret, package script, network target, or code-execution surface.

Review findings

  • [P1] Restore hosted-feed alignment for the fallback row — extensions/nvidia/openclaw.plugin.json:49
Review details

Best possible solution:

Keep the bundled fallback aligned with NVIDIA's hosted featured feed unless NVIDIA raises the hosted row or maintainers explicitly sponsor and prove a separate long-context fallback policy.

Do we have a high-confidence way to reproduce the issue?

Yes for the source-level merge risk: current main and the live NVIDIA feed use 262,144 for Super, while PR head changes only the bundled fallback to 1,048,576. There is no high-confidence user-flow proof that the hosted OpenClaw NVIDIA path should use the larger fallback limit today.

Is this the best way to solve the issue?

No. The edit is narrow, but changing only fallback metadata is not the best fix without maintainer sponsorship because it diverges from the hosted feed contract; alignment or an explicit long-context policy is safer.

Full review comments:

  • [P1] Restore hosted-feed alignment for the fallback row — extensions/nvidia/openclaw.plugin.json:49
    NVIDIA's current featured feed still returns 262,144 for Super, and current main maps that hosted value directly into contextWindow. Changing only the bundled fallback to 1,048,576 makes fallback users see a larger context budget than live-catalog users, so compaction and over-limit behavior can depend on feed availability unless maintainers explicitly sponsor that divergence.
    Confidence: 0.92

Overall correctness: patch is incorrect
Overall confidence: 0.92

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 477b27b6f87e.

Label changes

Label justifications:

  • P2: This is a bounded NVIDIA provider metadata change with real context-budget impact but limited provider-scoped blast radius.
  • merge-risk: 🚨 compatibility: Merging can change fallback context limits and compaction behavior for existing NVIDIA users when the hosted feed is unavailable or not used.
  • rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🧂 unranked krab.
  • status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The PR body lists validation commands but no current-head terminal output, logs, live output, or linked artifact showing the patched fallback behavior; posted proof should be redacted and then the PR body can be updated for re-review. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.
Evidence reviewed

PR surface:

Source 0, Tests +4, Docs 0. Total +4 across 3 files.

View PR surface stats
Area Files Added Removed Net
Source 1 1 1 0
Tests 1 4 0 +4
Docs 1 1 1 0
Config 0 0 0 0
Generated 0 0 0 0
Other 0 0 0 0
Total 3 6 2 +4

What I checked:

  • Repository review policy applied: Root policy requires current source, shipped/current behavior, tests/CI evidence, and dependency-backed proof for findings; this affected the review because provider fallback metadata is a compatibility-sensitive catalog surface. (AGENTS.md:39, 477b27b6f87e)
  • Scoped extension metadata policy applied: The extension guide says plugin metadata in openclaw.plugin.json must stay accurate so discovery and setup work without executing plugin code. (extensions/AGENTS.md:32, 477b27b6f87e)
  • Current main fallback value: Current main keeps the bundled Nemotron 3 Super fallback contextWindow at 262144. (extensions/nvidia/openclaw.plugin.json:49, 477b27b6f87e)
  • Current main live-feed contract: Current main maps NVIDIA featured-feed entry.context directly into contextWindow, so hosted feed values pass through unchanged. (extensions/nvidia/provider-catalog.ts:195, 477b27b6f87e)
  • Current docs contract: Current docs say setup/model-selection prefer NVIDIA's public featured catalog when available and fall back to the bundled catalog when unavailable; the bundled table lists Super at 262,144. Public docs: docs/providers/nvidia.md. (docs/providers/nvidia.md:70, 477b27b6f87e)
  • Live NVIDIA hosted feed: The public NVIDIA featured-model feed returned Nemotron 3 Super with context 262144 and max-output 8192 during review.

Likely related people:

  • eleqtrizit: Authored the merged NVIDIA live featured-catalog implementation that established the current live-feed plus bundled-fallback contract, and also authored this branch. (role: recent live-catalog contributor; confidence: high; commits: 6fd4aa8a2765; files: extensions/nvidia/provider-catalog.ts, extensions/nvidia/openclaw.plugin.json, docs/providers/nvidia.md)
  • jacobtomlinson: Changed the NVIDIA default to Ultra and touched the same fallback catalog, docs, and provider tests while preserving successful featured-feed authority. (role: recent area contributor; confidence: medium; commits: 829847292e22; files: extensions/nvidia/openclaw.plugin.json, extensions/nvidia/provider-catalog.ts, docs/providers/nvidia.md)
  • steipete: Closed the earlier same-change PR with the hosted-feed-authority rationale and has adjacent NVIDIA provider/docs history. (role: prior reviewer for same catalog decision; confidence: high; commits: 60f8e18372e0, 759965a31685; files: extensions/nvidia/provider-catalog.ts, docs/providers/nvidia.md)
What the crustacean ranks mean
  • 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
  • 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
  • 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
  • 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
  • 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
  • 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
  • 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works
  • ClawSweeper keeps one durable marker-backed review comment per issue or PR.
  • Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
  • A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
  • PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
  • Maintainers can also comment @clawsweeper review to request a fresh review only.
  • Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
  • Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
  • Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

@clawsweeper clawsweeper Bot added rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. status: 📣 needs proof The PR needs real behavior proof before ClawSweeper can clear the contributor ask. P2 Normal backlog priority with limited blast radius. merge-risk: 🚨 compatibility 🚨 May break existing users, config, migrations, defaults, or upgrade paths. labels Jul 1, 2026
@eleqtrizit
eleqtrizit merged commit fe43343 into openclaw:main Jul 1, 2026
91 of 94 checks passed
@github-actions github-actions Bot mentioned this pull request Jul 1, 2026
Open
liuhao1024 pushed a commit to liuhao1024/openclaw that referenced this pull request Jul 1, 2026
@eleqtrizit @liuhao1024
fix(nvidia): use Nemotron Super 1M context (openclaw#98726)
80959dc 
* fix(nvidia): correct Nemotron Super context
This was referenced Jul 2, 2026
Closed
Merged
Closed
github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request Jul 2, 2026
@eleqtrizit
fix(nvidia): use Nemotron Super 1M context (openclaw#98726)
51b2525 
* fix(nvidia): correct Nemotron Super context
@eleqtrizit eleqtrizit self-assigned this Sep 13, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

@eleqtrizit eleqtrizit

Labels

docs Improvements or additions to documentation extensions: nvidia maintainer Maintainer-authored PR merge-risk: 🚨 compatibility 🚨 May break existing users, config, migrations, defaults, or upgrade paths. P2 Normal backlog priority with limited blast radius. rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. size: XS status: 📣 needs proof The PR needs real behavior proof before ClawSweeper can clear the contributor ask.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@eleqtrizit