examples: add LlamaIndex check_action demo

[kkkhs]

What & why

Closes #204.

Adds a LlamaIndex refund_order tool example that runs check_action before the mock refund, wires it into the examples package, and covers allowed/blocked verdicts in the local examples test. The example imports tool from @llamaindex/core/tools instead of the top-level llamaindex package, so the examples package does not pull in LlamaIndex's workflow or parser/native dependency chains.

Test plan

• pnpm install --frozen-lockfile
• pnpm --filter @gnt-ai/examples lint
• pnpm --filter @gnt-ai/examples typecheck
• pnpm --filter @gnt-ai/examples build
• pnpm --filter @gnt-ai/examples test
• pnpm turbo run lint typecheck build --filter=@gnt-ai/examples
• pnpm audit --audit-level moderate
• git diff --check

Before you open this

Reviewers here hold PRs to the same bar the maintainers hold their own changes to. Check these
before requesting review, not after:

• Commits are signed off (git commit -s) — see CONTRIBUTING.md's DCO section.
• Functionally correct: you ran this, not just read it. If it touches apps/api or
  apps/store, you ran the relevant test suite locally.
• No dead code — no unused imports/variables, no commented-out blocks, no debug logging left
  behind.
• Non-trivial logic (a new branch, a parser, anything security- or money-adjacent) has a
  test. A one-line change doesn't need one; a new code path does.
• If this touches anything multi-tenant (rules, connectors, MCP keys, billing), it respects
  org/tenant isolation — no query or check that could leak one org's data to another.
• Matches this repo's existing patterns rather than introducing a new one for the same
  problem — see CONTRIBUTING.md's code conventions.
• If this touches apps/cli/src/prebrain/extraction/, the PR description includes the
  recall/precision numbers from bun run eval:extraction -- --mode cloud.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Team

Run ID: 51c22520-6e49-4d2c-8d6f-89f5ce877643

📥 Commits

Reviewing files that changed from the base of the PR and between 59a743d and d0b3ad2.

⛔ Files ignored due to path filters (1)

• pnpm-lock.yaml is excluded by !**/pnpm-lock.yaml

📒 Files selected for processing (5)

• examples/README.md
• examples/package.json
• examples/src/llamaindex-check-action.ts
• examples/test/guard.test.ts
• pnpm-workspace.yaml

🚧 Files skipped from review as they are similar to previous changes (5)

• examples/README.md
• pnpm-workspace.yaml
• examples/test/guard.test.ts
• examples/package.json
• examples/src/llamaindex-check-action.ts

Included review availability: Your plan provides up to 1 included review per hour; 0 remain after this review.

---

📝 Walkthrough

Walkthrough

Adds a LlamaIndex refund tool example that validates input, checks gnt policy before a mock refund, supports direct execution, and tests allowed and blocked verdicts. It also adds the package script, dependency, README command, and pnpm build-script setting.

Changes

LlamaIndex refund example

Layer / File(s)	Summary
Tool contract and guarded execution examples/src/llamaindex-check-action.ts	Defines validated refund inputs, calls checkAction, runs the refund through runGuardedAction, and returns a JSON-stringified result from the refund_order tool.
Runnable example wiring examples/src/llamaindex-check-action.ts, examples/package.json, examples/README.md, pnpm-workspace.yaml	Adds the LlamaIndex dependency and example script, documents the command, connects and closes the gnt MCP client, and disables tree-sitter build scripts.
Allowed and blocked verdict validation examples/test/guard.test.ts	Tests executed results for allowed verdicts and blocked results for denied verdicts without running the refund action.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Merge Risk: ⚪ Minimal · up to d0b3a

This PR adds a localized LlamaIndex refund-order example and related tests and package wiring; no actionable merge-blocking risk remains beyond normal checks and review.

Sequence Diagram(s)

sequenceDiagram
  participant runExample
  participant refund_order
  participant GntChecker
  participant runGuardedAction
  participant mockRefund
  runExample->>refund_order: execute orderId and amount
  refund_order->>GntChecker: checkAction(description, context)
  GntChecker-->>refund_order: policy verdict
  refund_order->>runGuardedAction: execute with verdict
  runGuardedAction->>mockRefund: run when allowed
  mockRefund-->>refund_order: refund result
  refund_order-->>runExample: JSON-stringified result

Loading

🚥 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%. Docstring coverage is scoped to functions touched by this diff. Analyzed 5 functions across 2 files. (3 skipped: 3 …	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	The PR satisfies issue #204 by adding the LlamaIndex refund tool example, wiring the example:llamaindex script, following the shared guard and MCP helpers, and adding allowed and blocked verdict tests…
Out of Scope Changes check	✅ Passed	The changes remain within the linked issue scope. The tree-sitter build configuration supports the new LlamaIndex dependency and does not introduce unrelated functionality.
Title check	✅ Passed	The title clearly and concisely identifies the main change: adding a LlamaIndex check_action demo.
Description check	✅ Passed	The description explains what changed, why it changed, the test plan, and the required checklist. It is complete and aligned with the pull request objectives.

Full details: Linked Issues check

Explanation

The PR satisfies issue #204 by adding the LlamaIndex refund tool example, wiring the example:llamaindex script, following the shared guard and MCP helpers, and adding allowed and blocked verdict tests.

Full details: Docstring Coverage

Explanation

Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 5 functions across 2 files. (3 skipped: 3 unsupported.)

• Fix all pre-merge checks with AI

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[lukaadzic]

@kkkhs the LlamaIndex demo is still marked draft. If it's ready, flip it out of draft and I'll review and run CI on it. #205 (Pydantic AI) and #229 just landed if you want a recently-merged example to line the shape up against.

Unrelated ask: I noticed you forked the repo but haven't starred it. If gnt has been useful to you, a star genuinely helps. It's most of how people find this thing, and the people who need a policy check in front of their agent mostly aren't searching for it yet.

[Copilot]

🟡 Changes recommended

The new llamaindex dependency pulls in deprecated workflow packages and native-parser transitive deps, and the PR should either narrow the dependency surface or document/justify the choice.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Adds a new LlamaIndex-based example to the @gnt-ai/examples package demonstrating a refund_order tool that performs check_action before executing a mock refund, aligning LlamaIndex with the existing framework examples (LangChain, Vercel AI SDK, OpenAI Agents).

Changes:

• Add examples/src/llamaindex-check-action.ts implementing a LlamaIndex tool that gates refund execution on check_action.
• Wire a new example:llamaindex script and document it in the examples README.
• Extend the examples test suite to cover allowed/blocked outcomes for the LlamaIndex tool and update the workspace to keep tree-sitter builds disabled.

File summaries

File	Description
pnpm-workspace.yaml	Disables tree-sitter builds in the workspace to avoid native compilation.
pnpm-lock.yaml	Adds/updates lock entries for the new llamaindex dependency and its transitive deps.
examples/test/guard.test.ts	Adds tests validating the LlamaIndex refund tool checks gnt and respects blocked verdicts.
examples/src/llamaindex-check-action.ts	New LlamaIndex refund_order tool example using check_action + runGuardedAction.
examples/README.md	Documents the new example:llamaindex command.
examples/package.json	Adds the example:llamaindex script and the llamaindex dependency.

Review details

Files not reviewed (1)

• pnpm-lock.yaml: Generated file

• Files reviewed: 5/6 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[kkkhs]

Done. I flipped the PR out of draft and narrowed the LlamaIndex import/dependency to @llamaindex/core/tools, so the broader llamaindex dependency and tree-sitter build config are gone.