Walkthrough

This PR performs a comprehensive code cleanup across the NemoClaw codebase, removing unused imports, helper functions, and constants from source and test files. It includes a significant refactoring of wiring in src/lib/onboard.ts, enhances permission-safety test setups with explicit chmod calls, improves test assertions using spy-based verification, and updates test size budgets accordingly.

Changes

Code Cleanup and Test Enhancement

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• NVIDIA/NemoClaw#4628: Updates to permission-healing test setup in src/lib/state/config-io.test.ts directly correspond to symlink-safety and chmod behavior covered in that PR.
• NVIDIA/NemoClaw#4905: Both PRs modify ci/test-file-size-budget.json—the retrieved PR introduces the budget guard, and this PR adjusts entries for tests with changed line counts.

Suggested labels

area: cli, NemoClaw CLI, Sandbox

Suggested reviewers

• prekshivyas
• cjagwani

🐰 Hops through the codebase with glee,
Removing unused imports, setting helpers free!
With chmod guards and spy-checks so fine,
The tests now behave—everything's in line! ✨

❌ Failed checks (1 warning)

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

E2E Recommendation Advisor

Base: origin/main
Head: HEAD
Confidence: medium

Required E2E

• None. No merge-blocking E2E is recommended: the production-file edits shown are unused import/unused symbol cleanup in critical onboarding and inference modules, while the remaining changes are tests or CI tooling. Build/type/unit coverage should catch accidental symbol issues, and there is no apparent runtime behavior change to installer, onboarding flow, sandbox lifecycle, credentials, network policy, or inference routing.

Optional E2E

• ollama-proxy-e2e (medium (~15 minutes; installs Ollama and pulls a small model)): Optional confidence because src/lib/inference/ollama/proxy.ts was touched. The diff looks like unused-code cleanup, but this existing E2E validates real Ollama auth proxy token auth, inference through the proxy, persistence, recovery, and container reachability.
• strict-tool-call-probe-e2e (low-medium): Optional confidence because src/lib/inference/onboard-probes.ts was touched. This hermetic regression E2E validates strict Chat Completions tool-call probe payload shape and retry behavior without needing GPU/Ollama infrastructure.
• onboard-inference-smoke-e2e (low-medium): Optional confidence because src/lib/onboard.ts and inference probe wiring were touched. This regression E2E checks that onboarding does not report success until the configured inference route serves a real request.

New E2E recommendations

• None.

E2E Scenario Advisor

Base: origin/main
Head: HEAD
Confidence: medium

Required scenario E2E

• ubuntu-repo-cloud-openclaw: Core onboarding and inference probe code changed in src/lib/onboard.ts and src/lib/inference/onboard-probes.ts; this is the primary Ubuntu repo-current cloud OpenClaw scenario covering baseline onboarding, sandbox creation, cloud inference, and credentials.
  • Dispatch: gh workflow run e2e-scenarios.yaml --ref <pr-head-ref> --field scenarios=ubuntu-repo-cloud-openclaw
• gpu-repo-local-ollama-openclaw: src/lib/inference/ollama/proxy.ts changed; the GPU local Ollama scenario is the only dispatchable scenario route that exercises the Ollama auth proxy/local Ollama onboarding surface.
  • Dispatch: gh workflow run e2e-scenarios.yaml --ref <pr-head-ref> --field scenarios=gpu-repo-local-ollama-openclaw

Optional scenario E2E

• ubuntu-repo-cloud-hermes: Optional adjacent coverage for the same repo-current cloud onboarding/inference path with the Hermes agent; useful because src/lib/onboard.ts is shared across agents, but OpenClaw is the primary target.
  • Dispatch: gh workflow run e2e-scenarios.yaml --ref <pr-head-ref> --field scenarios=ubuntu-repo-cloud-hermes
• ubuntu-repo-openai-compatible-openclaw: Optional adjacent coverage for OpenAI-compatible onboarding/probe selection paths touched by src/lib/inference/onboard-probes.ts; not primary because the diff appears to remove unused imports rather than alter probe behavior.
  • Dispatch: gh workflow run e2e-scenarios.yaml --ref <pr-head-ref> --field scenarios=ubuntu-repo-openai-compatible-openclaw

Relevant changed files

• src/lib/inference/ollama/proxy.ts
• src/lib/inference/onboard-probes.ts
• src/lib/onboard.ts
• test/e2e-scenario/framework-tests/e2e-scenario-registry.test.ts

• **Runtime validation** — Verify the source-shape scanner still reports production source reads derived through helper functions and source-tree enumeration after removing dead scanner helpers.. The patch touches onboarding/inference and infrastructure-adjacent files, but the actual production changes are unused-declaration removals. Existing focused tests cover scanner behavior, permission repair boundaries, and policy temp-dir ordering; runtime validation is useful for confidence but no specific missing test blocks this cleanup.
• **Runtime validation** — Verify `policy applyPreset` exits on an unresolved OpenShell binary before requesting a `nemoclaw-policy-*` temp directory.. The patch touches onboarding/inference and infrastructure-adjacent files, but the actual production changes are unused-declaration removals. Existing focused tests cover scanner behavior, permission repair boundaries, and policy temp-dir ordering; runtime validation is useful for confidence but no specific missing test blocks this cleanup.
• **Runtime validation** — Verify config permission tests preserve the symlink target mode and heal only host `~/.nemoclaw` siblings under a restrictive umask.. The patch touches onboarding/inference and infrastructure-adjacent files, but the actual production changes are unused-declaration removals. Existing focused tests cover scanner behavior, permission repair boundaries, and policy temp-dir ordering; runtime validation is useful for confidence but no specific missing test blocks this cleanup.