Elastic Peek: AI Software Engineering Factory (Research Project)

This repository is a proof-of-concept AI Software Engineering Factory built on elastic/ai-github-actions. It demonstrates how a suite of AI-powered GitHub Actions workflows can autonomously triage issues, review pull requests, propose fixes, and iterate on a real codebase with minimal human intervention.

Elastic Peek is both a way to look into your cluster and your data, and a look at the future of AI-powered software factories.

This is a research project, not an official Elastic product.

How it works

AI workflows in .github/workflows/ handle the software engineering lifecycle:

• Issue triage & duplicate detection — new issues are classified, labeled, and checked for duplicates
• Automated PR review — pull requests receive AI-generated code reviews with actionable feedback
• Fix proposals — the factory can draft code changes in response to issues
• Docs & text auditing — scheduled patrols check documentation quality and prose clarity
• Stale issue management — aging issues are flagged and closed automatically

The workflows are powered by elastic/ai-github-actions and run in standard GitHub Actions CI — no custom infrastructure required.

Copilot PAT setup (including private repositories)

These workflows require a repository secret named COPILOT_GITHUB_TOKEN.

1. Create a GitHub Personal Access Token (PAT) for the account that will run the workflows.
2. Grant the token access to the repository features your workflows use (for example: issues, pull requests, and contents).
3. Add it as a repository secret named COPILOT_GITHUB_TOKEN.

The same setup applies to private repositories; just make sure the PAT is authorized for the private repository (or org) where the workflows run.

Gemini API key setup

Workflows that use Google Gemini (Deep Research, Ideas agents) require a repository secret named GEMINI_API_KEY.

1. Obtain an API key from Google AI Studio.
2. Add it as a repository secret named GEMINI_API_KEY.

---

The Product: Elastic Peek

The codebase that the factory operates on is Elastic Peek, a lightweight, browser-only, backend-free UI for interacting with Elasticsearch. It's powered by Perses components and Elasticsearch ES|QL.

Live Demo

https://elastic.github.io/ai-github-actions-playground/

Overview

Elastic Peek is a browser-based dashboard builder that connects directly to your Elasticsearch cluster. The static site queries Elasticsearch using ES|QL via the _query REST API — or optionally via a local proxy to avoid CORS.

• Direct browser-to-Elasticsearch — or via a local proxy to avoid CORS
• ES|QL query editor — write queries with syntax highlighting
• No-code metrics explorer — browse, filter, and chart metrics from UI selections
• Recent query history — quickly re-use successful queries in Query Lab and Panel Editor
• Multiple visualization types — time series, bar charts, pie charts, tables, stats, gauges, markdown panels
• Drag-and-drop layout — resize and rearrange panels freely
• Persistent state — dashboards save to localStorage automatically
• Import/export — share dashboards as JSON files
• Dark and light themes

Prerequisites

• Node.js ^20.19.0 or >=22.12.0 (required by Vite 7)
• npm >=10 (bundled with Node.js 20/22)
• GNU Make — pre-installed on macOS and most Linux distributions. On Windows, install via Chocolatey (choco install make), Scoop (scoop install make), or use WSL.

Use a version manager such as nvm or fnm to switch Node versions quickly. A .nvmrc file is included in this repo — run nvm use or fnm use at the repo root to activate the correct version automatically.

See DEVELOPING.md for the full development guide.

Quick Start

make setup   # install dependencies
make serve   # start dev server at http://localhost:3000

If make is unavailable, run the equivalent npm commands directly:

cd peek
npm ci        # install dependencies
npm run dev   # start dev server at http://localhost:3000

To avoid CORS, use the built-in proxy or Electron mode — see DEVELOPING.md for details.

ES_URL=http://localhost:9200 make serve-proxy  # proxy mode
make electron-dev                               # native desktop app (no CORS needed)

In-Product Docs

Use the Docs tab in the app for embedded documentation and search covering getting started, connecting, CORS/proxy setup, Query Lab, Metrics, Traces, Profiling, dashboard workflows, Console, Data Streams, Chat, LLM settings, Cluster Overview, Users & Roles, and dashboard management.

Further Reading

• Development guide
• Contributing
• Documentation standards

License

MIT