LarkOps AI turns broken CI into verified fixes.
It is a polished hackathon demo for the Lark sponsor challenge at DevNetwork AI + ML Hackathon 2026. The product presents an autonomous debugging workflow where Lark CLI/MCP acts as the execution and replay layer for failure investigations.
- Landing page with the product story and "Launch Demo Investigation" path
- Command-center dashboard for recent software failures
- New investigation flow for CI logs, failing tests, branches, commits, and issue context
- Detailed investigation page with Lark workflow timeline, terminal logs, agents, patch, regression test, verification, confidence gate, PR preview, failure memory, Lark events, and postmortem export
- Settings page for Lark CLI/MCP status and Gemini status
- Deterministic demo mode when no API keys are configured
Lark is central to the product story. The UI exposes:
- Lark workflow status
- Lark MCP workflow mode
- Lark replay link
- Lark event IDs
- Lark execution timeline
- Lark-powered reproduction step
- Lark usage events
- Lark investigation metadata
- Real Lark CLI workflow creation and invocation through
GETLARK_API_KEY
The verified real workflow is:
npx -y @getlark/cli workflows invoke --workflow-ids wflw_sEkzdOqOLAutmd5dXsbxUkua --wait --verboseThe app still has deterministic demo outputs so judges can experience the product without credentials, but this workspace has also been verified against a real Lark CLI workflow.
Gemini is the default AI provider. If VITE_GEMINI_API_KEY is present, the app labels itself as Gemini mode. Without it, the app runs in deterministic demo/mock mode and still produces high-quality investigation outputs.
npm install
npm run devThen open the local Vite URL, usually:
http://127.0.0.1:5173/
If that port is busy, Vite will print the next available port.
npm run build
npm run lintCreate .env.local if you want the UI to show Gemini mode:
VITE_GEMINI_API_KEY=your_gemini_key_here
VITE_GEMINI_MODEL=gemini-2.0-flash
VITE_LARK_WORKFLOW_ID=wflw_sEkzdOqOLAutmd5dXsbxUkua
VITE_LARK_WORKFLOW_STATUS=success
VITE_LARK_EXECUTION_ID=wflw_exec_i8dNNfBsn0p1gvgkIgA8wwdp
VITE_LARK_CLI_VERIFIED=true
GETLARK_API_KEY=your_lark_key_hereReal Lark credentials are not required for the demo path, but GETLARK_API_KEY enables CLI-backed workflow verification.
The current demo workflow was created, generated, and invoked with the real Lark CLI.
Workflow ID: wflw_sEkzdOqOLAutmd5dXsbxUkua
Last execution ID: wflw_exec_i8dNNfBsn0p1gvgkIgA8wwdp
Last execution result: success