fix: sync OpenClaw main agent model (Fixes #5822)

[deepujain]

Summary

Fixes #5822.

nemoclaw inference set already updated the OpenShell gateway route, the registry, and agents.defaults.model.primary in /sandbox/.openclaw/openclaw.json. It missed an explicit agents.list entry for the primary main agent when that entry carried its own model, so OpenClaw could keep reporting the old model even after status and the gateway had moved to the new route.

This patch updates the explicit primary agent model during the same sandbox config sync, while leaving secondary agent model overrides alone.

Changes

• src/lib/actions/inference-set.ts: sync agents.list entries for id: "main" or default: true when they already have a model.
• src/lib/actions/inference-set.test.ts: add regression coverage for the stale primary-agent model path.

Testing

• npx vitest run src/lib/actions/inference-set.test.ts - 27 tests passed.
• npm run build:cli - passed.
• npm run typecheck:cli - passed.
• npm run source-shape:check - passed outside the sandbox after tsx hit a sandbox IPC permission error on its temp pipe.
• npx @biomejs/biome format src/lib/actions/inference-set.ts src/lib/actions/inference-set.test.ts - passed.

Evidence it works

The new regression test builds an OpenClaw config where agents.defaults.model.primary and agents.list[id=main].model both point at the old model. After patchOpenClawInferenceConfig(...), both primary identity fields point at the switched model, while the secondary agent model remains unchanged. Before this patch, the explicit main model stayed stale.

I did not run a live DGX Spark/OpenClaw TUI repro in this worktree, so this PR covers the config-sync regression with focused unit evidence rather than a full hardware-backed TUI smoke.

Signed-off-by: Deepak Jain deepujain@gmail.com

Summary by CodeRabbit

• Bug Fixes
  • Updating the primary agent model now also updates the corresponding primary entry in the agent list to keep the selected model consistent.
  • Added improved handling when no dedicated main/default entry exists, ensuring the default primary agent is updated correctly.
  • Non-primary agent entries keep their existing model and details.
• Tests
  • Added test coverage for both the main-entry update scenario and the fallback scenario when patching inference configuration.

[copy-pr-bot]

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.

[coderabbitai]

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

• @coderabbitai resume to resume automatic reviews.
• @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

• ▶️ Resume reviews
• 🔍 Trigger review

📝 Walkthrough

Walkthrough

updateAgentPrimary now also syncs the primary model into agents.list, preferring the main entry and falling back to the first default entry. New tests cover both the main-entry and fallback update paths.

Changes

OpenClaw primary model sync

Layer / File(s)	Summary
Primary model propagation to agents.list src/lib/actions/inference-set.ts	updateAgentPrimary now calls updatePrimaryAgentListModel, which updates the primary string model in agents.list, preferring an id === "main" entry and falling back to a default === true entry.
Tests for main-entry and fallback sync src/lib/actions/inference-set-patch-openclaw.test.ts	New Vitest cases verify patchOpenClawInferenceConfig updates the main agents.list entry when present, and falls back to the default: true entry when no main entry exists.

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested reviewers: jyaunches, ericksoa, prekshivyas

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely summarizes the main change: syncing the OpenClaw main agent model.
Linked Issues check	✅ Passed	The change updates the primary main agent model while preserving secondary overrides, matching the regression fix requested in #5822.
Out of Scope Changes check	✅ Passed	The added test coverage and model-sync helper stay within the issue scope and do not introduce unrelated behavior.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

src/lib/actions/inference-set.ts (1)

225-242: 🩺 Stability & Availability | 🔵 Trivial

Please run the targeted inference-switch E2Es for this path.

This code path patches the running agent config during inference switching, so it is worth validating both the OpenClaw and Hermes flows before merge. As per path instructions, src/lib/actions/inference-set.ts “affect[s] OpenClaw and Hermes model switching without a sandbox rebuild,” and the recommended jobs are openclaw-inference-switch-e2e and hermes-inference-switch-e2e.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/lib/actions/inference-set.ts` around lines 225 - 242, Run the targeted
inference-switch E2Es for the config-patching path in
updateAgentPrimary/updatePrimaryAgentListModel, since this change affects
OpenClaw and Hermes model switching without a sandbox rebuild. Validate both
flows before merge by running the openclaw-inference-switch-e2e and
hermes-inference-switch-e2e jobs to confirm the running agent config is updated
correctly during inference switching.

Source: Path instructions

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/lib/actions/inference-set.ts`:
- Around line 236-240: The primary-model rewrite in the list loop is too broad
because it treats any entry with default: true as primary. Update the logic in
inference-set so id === "main" is matched and rewritten first, and only use a
default-flagged entry if no main entry exists, ensuring secondary overrides
marked default: true are left unchanged. Focus the fix around the list iteration
and the isPrimaryAgent check that sets entry.model to primaryModelRef.

---

Nitpick comments:
In `@src/lib/actions/inference-set.ts`:
- Around line 225-242: Run the targeted inference-switch E2Es for the
config-patching path in updateAgentPrimary/updatePrimaryAgentListModel, since
this change affects OpenClaw and Hermes model switching without a sandbox
rebuild. Validate both flows before merge by running the
openclaw-inference-switch-e2e and hermes-inference-switch-e2e jobs to confirm
the running agent config is updated correctly during inference switching.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: a7490e22-019a-4928-ad0c-0fb534e7b8f2

📥 Commits

Reviewing files that changed from the base of the PR and between 06485be and 9d793bb.

📒 Files selected for processing (2)

• src/lib/actions/inference-set.test.ts
• src/lib/actions/inference-set.ts

[wscurran]

✨ Thanks for syncing the explicit agents.list entry for the primary main agent during the sandbox config sync. This proposes a way to update the OpenClaw main agent model when nemoclaw inference set switches the gateway route while preserving secondary agent model overrides.

---

Related open issues:

• #5822 [All Platforms][Inference][GitHub Issue #5822] NemoClaw v0.0.67: nemoclaw inference set switches gateway model but OpenClaw agent still uses old model (6154913 regression)

[deepujain]

Rebased on current main and moved the regression coverage into the new focused OpenClaw inference patch test. Build, typecheck, lint/checks, and focused inference Vitest all pass. Ready for another look.

[deepujain]

Rebased on current main again. Build, typecheck, lint/checks, and focused OpenClaw inference tests pass; GitHub checks are green and the commit is verified.

[deepujain]

Rebased on current main and refreshed the verified commit. Build, typecheck, lint/checks, conditional scan, and focused OpenClaw inference tests pass.

[deepujain]

Rebased on current main and refreshed the verified commit. Build, typecheck, lint/checks, conditional scan, and focused OpenClaw inference tests pass.

[cv]

Reviewed the current-main head. The inference update synchronizes an explicit main agent model (or the default agent when no main entry exists) without mutating secondary agent models, with focused regression coverage. All current-head CI and contributor-compliance gates are green.

[prekshivyas]

Advisor follow-up / maintainer disposition:

• The GPT-5.5 advisor recommends merge_as_is with no required findings.
• Nemotron's malformed-agents.list warning is not applicable as a correctness or security blocker. agents.list is optional in valid OpenClaw configurations, so absence is a normal case and should not emit a warning during every inference switch. The updater intentionally mutates only an existing string model; it does not invent a field or coerce an invalid non-string value. agents.defaults.model.primary is still updated in every valid call, so the supported fallback remains authoritative.
• The explicit id: "main" and default-agent paths are both covered, including preservation of a secondary default: true override. CodeRabbit's earlier functional concern is resolved.
• The architecture/JSDoc and higher-level mocked persistence items are non-blocking suggestions; the changed helper is pure, and the full CLI suite plus focused patch tests are green.
• The E2E target advisor requires openclaw-inference-switch. It is dispatched against the exact PR SHA in run https://github.com/NVIDIA/NemoClaw/actions/runs/28929454796 and will be tracked to completion.

On this evidence, I accept the current defensive no-op behavior for missing or malformed optional list entries and do not want user-facing noise added for valid configs without agents.list.

[github-actions]

E2E Target Results — ✅ All requested jobs passed

Run: 28929454796
Workflow ref: ci/pr-5857-prekshivyas
Requested targets: (default — all supported)
Requested jobs: openclaw-inference-switch
Summary: 1 passed, 0 failed, 0 cancelled, 0 skipped

Job	Result
openclaw-inference-switch	✅ success

[prekshivyas]

Reviewed the config mutation and regression coverage. The canonical main agent is updated first, fallback only applies when no main entry exists, secondary overrides are preserved, standard CI is green, GPT-5.5 recommends merge-as-is, the Nemotron warning is explicitly dispositioned for optional agents.list behavior, and the required openclaw-inference-switch E2E passed in run 28929454796.