chore(cubebot): GitHub automation bot for issue triage

.github/workflows/deploy-cubebot.yml

@@ -0,0 +1,29 @@
+name: Deploy CubeBot
+
+on:
+  push:
+    branches:
+      - main
+    paths:
+      - 'apps/cubebot/**'
+  pull_request:
+    branches:
+      - main
+    paths:
+      - 'apps/cubebot/**'
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    name: Deploy to Cloudflare Workers
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - uses: ./.github/actions/setup-node
+
+      - name: Deploy
+        run: pnpm --filter cubebot run deploy
+        env:
+          CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}
+          CLOUDFLARE_ACCOUNT_ID: ${{ secrets.CLOUDFLARE_ACCOUNT_ID }}

apps/cubebot/.gitignore

@@ -0,0 +1,3 @@
+data/
+.wrangler/
+.dev.vars

apps/cubebot/CLAUDE.md

@@ -0,0 +1,33 @@
+# CubeBot
+
+GitHub App bot for TresJS issue triage built on Cloudflare Workers.
+
+## Tech Stack
+
+- **Runtime:** Cloudflare Workers
+- **Framework:** Hono
+- **Database:** Cloudflare D1
+- **AI:** Anthropic Claude + Workers AI (embeddings)
+- **GitHub:** Octokit
+
+## Key Files
+
+- `src/index.ts` - Worker entry point, webhook routing
+- `src/github/handlers/` - Event handlers for issues and comments
+- `src/ai/` - Claude integration and RAG retrieval
+- `src/triage/` - Issue detection and comment formatting
+- `scripts/embed-docs.ts` - Documentation embedding script
+
+## Testing Locally
+
+```bash
+pnpm dev
+# Expose via ngrok for webhook testing
+```
+
+## Secrets Required
+
+- `GITHUB_APP_ID`
+- `GITHUB_PRIVATE_KEY`
+- `GITHUB_WEBHOOK_SECRET`
+- `ANTHROPIC_API_KEY`

apps/cubebot/README.md

@@ -0,0 +1,98 @@
+# CubeBot 🧊
+
+GitHub App bot for TresJS issue triage and documentation assistance.
+
+## Features
+
+- Auto-triages new issues (bug vs feature detection)
+- Redirects feature requests to GitHub Discussions
+- Checks for required bug report info (reproduction, system info)
+- Suggests relevant documentation via RAG
+- Responds to @mentions with doc-based assistance
+
+## Setup
+
+### 1. Create GitHub App
+
+1. Go to github.com/settings/apps → New GitHub App
+2. Name: `tresjs-cubebot`
+3. Permissions:
+   - Issues: Read & Write
+   - Discussions: Read & Write
+4. Subscribe to events: `issues`, `issue_comment`
+5. Generate and download private key (`.pem` file)
+
+### 2. Cloudflare Setup
+
+```bash
+# Create D1 database
+pnpm wrangler d1 create cubebot-docs
+
+# Copy the database_id from output and update wrangler.toml
+
+# Apply schema to remote D1
+cd apps/cubebot
+pnpm wrangler d1 execute cubebot-docs --remote --file=schema.sql
+```
+
+### 3. Deploy Worker
+
+```bash
+cd apps/cubebot
+pnpm wrangler deploy
+```
+
+### 4. Set Secrets
+
+```bash
+cd apps/cubebot
+
+# App ID from GitHub App page
+pnpm wrangler secret put GITHUB_APP_ID
+
+# Contents of the .pem private key file
+pnpm wrangler secret put GITHUB_PRIVATE_KEY
+
+# Generate with: openssl rand -hex 32
+# Use same value in GitHub App webhook settings
+pnpm wrangler secret put GITHUB_WEBHOOK_SECRET
+
+# From console.anthropic.com
+pnpm wrangler secret put ANTHROPIC_API_KEY
+```
+
+### 5. Configure GitHub App Webhook
+
+1. Go to your GitHub App settings
+2. Set Webhook URL: `https://tresjs-cubebot.<account>.workers.dev/webhook`
+3. Set Webhook secret: same value as `GITHUB_WEBHOOK_SECRET`
+4. Save
+
+### 6. Seed Documentation (RAG)
+
+```bash
+# Call the admin endpoint to fetch docs and generate embeddings
+curl -X POST https://tresjs-cubebot.<account>.workers.dev/admin/seed-docs
+```
+
+This fetches `llms-full.txt` from TresJS doc sites, generates embeddings via Workers AI, and stores them in D1.
+
+### 7. Install GitHub App
+
+Install the app on the TresJS organization or specific repositories.
+
+## Development
+
+```bash
+cd apps/cubebot
+
+# Start local dev server
+pnpm wrangler dev
+
+# Use ngrok to expose for webhook testing
+ngrok http 8787
+```
+
+## Architecture
+
+See [design doc](../../.claude/plans/2026-01-27-cubebot-design.md) for full architecture details.

apps/cubebot/eslint.config.js

@@ -0,0 +1,5 @@
+import { tresLintConfig } from '@tresjs/eslint-config'
+
+export default tresLintConfig({
+  ignores: ['**/*.md'],
+})

apps/cubebot/package.json

@@ -0,0 +1,26 @@
+{
+  "name": "cubebot",
+  "type": "module",
+  "version": "0.0.1",
+  "private": true,
+  "scripts": {
+    "dev": "wrangler dev",
+    "deploy": "wrangler deploy",
+    "lint": "eslint .",
+    "lint:fix": "eslint . --fix",
+    "typecheck": "tsc --noEmit",
+    "embed-docs": "npx tsx scripts/embed-docs.ts"
+  },
+  "dependencies": {
+    "@anthropic-ai/sdk": "^0.78.0",
+    "@octokit/auth-app": "^8.2.0",
+    "hono": "^4.12.3",
+    "octokit": "^5.0.5"
+  },
+  "devDependencies": {
+    "@cloudflare/workers-types": "^4.20260226.1",
+    "@tresjs/eslint-config": "workspace:^",
+    "typescript": "catalog:typescript",
+    "wrangler": "^4.69.0"
+  }
+}

apps/cubebot/schema.sql

@@ -0,0 +1,11 @@
+CREATE TABLE IF NOT EXISTS doc_chunks (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  source TEXT NOT NULL,
+  url TEXT NOT NULL,
+  title TEXT NOT NULL,
+  content TEXT NOT NULL,
+  embedding TEXT NOT NULL,
+  created_at TEXT DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE INDEX IF NOT EXISTS idx_doc_chunks_source ON doc_chunks(source);

apps/cubebot/scripts/embed-docs.ts

@@ -0,0 +1,92 @@
+import { execSync } from 'node:child_process'
+
+const DOC_SOURCES = [
+  { name: 'core', url: 'https://docs.tresjs.org/llms-full.txt' },
+  { name: 'cientos', url: 'https://cientos.tresjs.org/llms-full.txt' },
+  { name: 'postprocessing', url: 'https://post-processing.tresjs.org/llms-full.txt' },
+]
+
+interface DocChunk {
+  source: string
+  url: string
+  title: string
+  content: string
+}
+
+async function fetchDocs(source: { name: string; url: string }): Promise<DocChunk[]> {
+  console.log(`Fetching ${source.name} docs from ${source.url}...`)
+
+  try {
+    const response = await fetch(source.url)
+    if (!response.ok) {
+      console.warn(`Failed to fetch ${source.url}: ${response.status}`)
+      return []
+    }
+
+    const text = await response.text()
+    return parseDocsIntoChunks(source.name, source.url, text)
+  }
+  catch (error) {
+    console.warn(`Error fetching ${source.url}:`, error)
+    return []
+  }
+}
+
+function parseDocsIntoChunks(source: string, baseUrl: string, text: string): DocChunk[] {
+  const chunks: DocChunk[] = []
+
+  // Split by headers (## or #)
+  const sections = text.split(/(?=^#{1,2}\s)/m)
+
+  for (const section of sections) {
+    if (section.trim().length < 50) continue // Skip tiny sections
+
+    const titleMatch = section.match(/^#{1,2}\s+(.+)$/m)
+    const title = titleMatch?.[1] ?? 'Untitled'
+
+    // Extract URL if present in the section
+    const urlMatch = section.match(/https?:\/\/[^\s)]+/)
+    const url = urlMatch?.[0] ?? baseUrl
+
+    chunks.push({
+      source,
+      url,
+      title: title.trim(),
+      content: section.slice(0, 2000), // Limit chunk size
+    })
+  }
+
+  return chunks
+}
+
+async function main() {
+  console.log('Starting docs embedding process...\n')
+
+  const allChunks: DocChunk[] = []
+
+  for (const source of DOC_SOURCES) {
+    const chunks = await fetchDocs(source)
+    allChunks.push(...chunks)
+    console.log(`  → ${chunks.length} chunks from ${source.name}`)
+  }
+
+  console.log(`\nTotal chunks: ${allChunks.length}`)
+
+  // Write chunks to a JSON file for manual D1 import
+  // In production, this would call Workers AI for embeddings
+  const outputPath = 'apps/cubebot/data/doc-chunks.json'
+
+  // Create data directory
+  execSync('mkdir -p apps/cubebot/data')
+
+  const fs = await import('node:fs')
+  fs.writeFileSync(outputPath, JSON.stringify(allChunks, null, 2))
+
+  console.log(`\nChunks written to ${outputPath}`)
+  console.log('\nNext steps:')
+  console.log('1. Run wrangler d1 execute to create the database')
+  console.log('2. Use a separate script to generate embeddings via Workers AI')
+  console.log('3. Import the embeddings into D1')
+}
+
+main().catch(console.error)