RFC: AI Documentation Directory Structure Proposal

This proposal outlines a modular AI documentation architecture:
- AGENTS.md as central hub
- Modular .ai/ guideline files
- Claude Code settings.json configuration
- Team vs personal configuration separation

Requesting team review and feedback on proposed structure.

Author: sg00dwin

PROPOSED-AI-DOCS-STRUCTURE.md

@@ -0,0 +1,350 @@
+# Proposed Directory Structure for AI Context & Configuration
+
+**Purpose**: Establish modular AI documentation and Claude Code configuration
+**Date**: 2025-11-25
+**Status**: Awaiting team approval
+
+---
+
+## Summary of Changes
+
+### What's Being Added
+- ✅ **AGENTS.md** - Central AI documentation hub
+- ✅ **.ai/README.md** - Explains .ai/ directory
+- ✅ **.ai/ARCHITECTURE.md** - System architecture, Plugin SDK, tech stack
+- ✅ **.ai/CONVENTIONS.md** - Coding standards, patterns, file naming
+- ✅ **.ai/TESTING.md** - All testing (unit, integration, E2E)
+- ✅ **.claude/settings.json** - Team Claude Code configuration
+
+### What's Being Updated
+- 🔄 **CLAUDE.md** - Update to reference AGENTS.md
+- 🔄 **.cursor/context.md** - Update to reference AGENTS.md
+
+### What's Being Removed
+- ❌ **.ai/context.md** - Content merged into new modular files
+
+---
+
+## Current Structure
+
+```
+console/
+├── CLAUDE.md                          # Claude Code entry, references .ai/context.md
+├── .ai/
+│   ├── context.md                     # Single monolithic file (will be deleted)
+│   └── README.md
+├── .claude/
+│   └── commands/                      # Team commands
+├── README.md
+├── CONTRIBUTING.md
+├── STYLEGUIDE.md
+└── INTERNATIONALIZATION.md
+```
+
+**Issues with Current Structure**:
+- ❌ No team `.claude/settings.json` (causes "create settings.json" suggestions)
+- ❌ Monolithic `.ai/context.md` (hard to maintain, update specific sections)
+- ❌ No central AI documentation hub
+- ❌ Mixed concerns in single file
+
+---
+
+## Proposed Structure
+
+```
+console/
+├── AGENTS.md                           # ✅ NEW - Central AI hub
+├── CLAUDE.md                           # 🔄 UPDATED - Points to AGENTS.md
+├── .ai/                                # Team AI context (AI-agnostic)
+│   ├──[context.md deleted]            # ❌ REMOVED - Content merged above
+│   ├── README.md                       # ✅ NEW - Explains .ai/ structure
+│   ├── ARCHITECTURE.md                 # ✅ NEW - System arch, Plugin SDK, tech stack
+│   ├── CONVENTIONS.md                  # ✅ NEW - Coding standards, patterns, etc.
+│   └── TESTING.md                      # ✅ NEW - All testing approaches
+├── .claude/                            # Claude Code configuration
+│   ├── settings.json                   # ✅ NEW - Team config (checked in)
+│   └── commands/                       # Team commands
+├── .cursor/                            # ✔️ EXISTS - Cursor-specific configs
+│   └── context.md                      # 🔄 UPDATED - Points to AGENTS.md
+├── README.md                           # ✔️ UNCHANGED
+├── CONTRIBUTING.md                     # ✔️ UNCHANGED
+├── STYLEGUIDE.md                       # ✔️ UNCHANGED
+├── INTERNATIONALIZATION.md             # ✔️ UNCHANGED
+└── coderabbitai.json                   # ✅ NEW - references .ai/ structure
+ 
+
+Legend:
+  ✅ NEW - File to be created
+  🔄 UPDATED - File to be modified
+  ❌ REMOVED - File to be deleted
+  ✔️ UNCHANGED - File stays as-is
+```
+
+---
+
+## File Purposes & Sizes
+
+### Root Level
+
+| File | Purpose | Size | Audience |
+|------|---------|------|----------|
+| **AGENTS.md** | Central AI documentation hub, quick start | 6-8KB | All AI tools |
+| **CLAUDE.md** | Claude Code entry point → AGENTS.md | Updated | Claude Code |
+
+### .ai/ Directory (Team AI Context)
+
+| File | Purpose | Size | Checked In |
+|------|---------|------|------------|
+| **README.md** | Explains .ai/ organization | 1-2KB | ✅ Yes |
+| **ARCHITECTURE.md** | System architecture, Plugin SDK, tech stack, monorepo | 10-12KB | ✅ Yes |
+| **CONVENTIONS.md** | Coding standards, P0/P1 patterns, file naming | 10-12KB | ✅ Yes |
+| **TESTING.md** | Unit, integration, E2E testing patterns | 6-8KB | ✅ Yes |
+
+**Total .ai/ documentation**: ~30-35KB (vs current 5KB context.md)
+
+### .claude/ Directory (Claude Code Configuration)
+
+| File | Purpose | Checked In |
+|------|---------|------------|
+| **settings.json** | Team permissions, hooks | ✅ Yes |
+| **commands/** | Shared slash commands | ✅ Yes |
+
+---
+
+## Content Distribution
+
+### .ai/ARCHITECTURE.md
+**Contains:**
+- Monorepo package structure (frontend/, pkg/, cmd/)
+- Technology stack overview:
+  - Frontend: React, TypeScript, yarn workspaces, Webpack Module Federation
+  - Backend: Go, klog, Kubernetes client libraries
+  - Deployment: OpenShift/Kubernetes
+  - Build tools: yarn, Go toolchain
+- Console Dynamic Plugin SDK architecture
+  - Extension points system (25+ types)
+  - Module Federation runtime loading
+  - Type system and code references
+  - Public API surface (re-exports from @console/shared, @console/internal, etc.)
+- Plugin structure patterns
+- Key architectural decisions
+- Package relationships and dependencies
+
+### .ai/CONVENTIONS.md
+**Contains:**
+- TypeScript/React conventions
+  - Functional components and hooks
+  - State management patterns (Context API, migrating from Redux)
+  - Component structure and file organization
+- Framework usage patterns:
+  - PatternFly design system usage
+  - React hooks best practices (useK8sWatchResource, etc.)
+  - i18n with useTranslation hook
+- Go best practices
+  - Package organization patterns
+  - Error handling standards
+  - Logging conventions (klog)
+  - HTTP handler patterns
+- File naming conventions (PascalCase, kebab-case rules)
+- API call patterns (consoleFetchJSON, k8s resource hooks)
+- Styling conventions (SCSS modules, PatternFly integration)
+- Error handling patterns
+- P0/P1 review patterns
+
+### .ai/TESTING.md
+**Contains:**
+- Unit testing patterns (Jest)
+- Integration testing approaches
+- E2E testing (Cypress)
+- Test organization and structure
+- Coverage expectations
+- Testing best practices
+
+---
+
+## How It Works
+
+### Session Loading Order
+
+When Claude Code starts a session:
+
+1. **CLAUDE.md** → Points to AGENTS.md
+2. **settings.json hook** → Loads AGENTS.md
+3. **settings.json hook** → Loads all .ai/ files
+4. **settings.local.json hook** → Loads personal workflow (if configured)
+
+### Team vs Personal Separation
+
+**Team Configuration** (checked into git, affects everyone):
+- `.claude/settings.json` - Team permissions and hooks
+- `.ai/` directory - All team guidelines and rules
+- `.claude/commands/` - Shared slash commands
+- `AGENTS.md` - Central documentation hub
+
+**Personal Configuration** (gitignored, individual developers, not checked in):
+- `.claude/settings.local.json` - Personal overrides and hooks
+- `.claude/local/` - Personal rules, preferences, analysis files
+
+### AI Tool Compatibility
+
+**Works With**:
+- ✅ Claude Code (via settings.json hooks)
+- ✅ CodeRabbit (via .coderabbit.yaml references)
+- ✅ Cursor/Copilot (via .cursorrules or similar)
+- ✅ Any AI tool (can read .ai/ files directly)
+
+---
+
+## Benefits
+
+### For Developers
+✅ **Clear separation**: Team standards vs personal workflow
+✅ **No conflicts**: Personal settings don't affect teammates
+✅ **No suggestions**: Claude stops suggesting "create settings.json"
+✅ **Better onboarding**: AGENTS.md provides quick start
+
+### For Maintainers
+✅ **Modular updates**: Update only relevant files when things change
+✅ **Scalable**: Easy to add new guideline files (ACCESSIBILITY.md, etc.)
+✅ **Version control**: Track changes to specific guidelines
+✅ **AI-agnostic**: Works with all AI tools, not just Claude
+
+---
+
+## What Gets Checked Into Git
+
+**Checked In** (team files):
+```
+✅ AGENTS.md
+✅ CLAUDE.md (updated)
+✅ .cursor/context.md (updated)
+✅ .ai/README.md
+✅ .ai/ARCHITECTURE.md
+✅ .ai/CONVENTIONS.md
+✅ .ai/TESTING.md
+✅ .claude/settings.json
+✅ .claude/commands/**
+✅ .gitignore (updated)
+```
+
+**Deleted**:
+```
+❌ .ai/context.md (content merged into new files)
+```
+
+---
+
+## Impact Assessment
+
+### Who is affected?
+- **All developers**: Will benefit from AI documentation
+- **Claude Code users**: Will benefit from settings.json configuration
+- **Cursor/Copilot users**: Can reference .ai/ files
+- **PR reviewers**: AI agents will follow established patterns
+
+### Breaking Changes?
+- ❌ **No breaking changes**
+- ✅ All existing files remain (except context.md merged)
+- ✅ Personal configurations (.claude/local/) unaffected
+- ✅ Backward compatible
+
+### Migration Required?
+- ❌ **No migration needed**
+- ✅ .ai/context.md content merged into new files
+- ✅ Existing workflows continue working
+- ✅ Claude Code benefits immediately from settings.json
+
+---
+
+## Implementation Timeline
+
+### Phase 0: Configuration (Day 1)
+- Create `.claude/settings.json`
+- Update `.gitignore`
+- Create `.ai/README.md`
+- Update `CLAUDE.md`
+
+### Phases 1-5: Documentation (Days 2-4)
+- Create `AGENTS.md`
+- Create `.ai/ARCHITECTURE.md` (system arch + tech stack)
+- Create `.ai/CONVENTIONS.md` (standards + patterns)
+- Create `.ai/TESTING.md` (all testing types)
+
+### Phase 6-7: Cleanup & Validation (Day 5)
+- Delete `.ai/context.md`
+- Validate all cross-references
+- Test with Claude Code session restart
+
+**Total**: ~5 days implementation time
+
+---
+
+## Questions for Team Review
+
+1. **Approve directory structure?** ✅ / ❌
+2. **Approve checking in `.claude/settings.json`?** ✅ / ❌
+3. **Approve deleting `.ai/context.md`?** ✅ / ❌
+4. **Any additional files needed?** _____________
+5. **Any concerns with gitignore strategy?** _____________
+6. **Preferred implementation timeline?** _____________
+
+---
+
+## Sample: .claude/settings.json (Team Configuration)
+
+This is what will be checked into git:
+
+```json
+{
+  "hooks": {
+    "SessionStart": [
+      {
+        "hooks": [
+          {
+            "type": "command",
+            "comment": "Load central AI guide",
+            "command": "cat AGENTS.md"
+          },
+          {
+            "type": "command",
+            "comment": "Load all team guidelines (3 .ai/ files)",
+            "command": "find .ai -type f -name '*.md' -not -name 'README.md' -exec cat {} +"
+          }
+        ]
+      }
+    ]
+  },
+  "permissions": {
+    "allow": [
+      "Read(//AGENTS.md)",
+      "Read(//CLAUDE.md)",
+      "Read(//.ai/**)",
+      "Read(//STYLEGUIDE.md)",
+      "Read(//INTERNATIONALIZATION.md)",
+      "Read(//CONTRIBUTING.md)",
+      "Read(//.claude/commands/**)",
+      "Bash(yarn test:*)",
+      "Bash(yarn build)",
+      "Bash(yarn lint)",
+      "Bash(yarn i18n)",
+      "Bash(./build-backend.sh)",
+      "Bash(./test-backend.sh)",
+      "Bash(git status)",
+      "Bash(git diff:*)",
+      "Bash(git log:*)",
+      "SlashCommand(/init)",
+      "SlashCommand(/plugin-api-review)"
+    ]
+  }
+}
+```
+
+**Why These Permissions?**
+- Common Console development operations
+- Safe read operations
+- Standard build/test commands
+- No destructive operations without asking
+
+**Personal Permissions?**
+- Developers can add more in `.claude/settings.local.json` (gitignored)
+- Personal settings extend (not override) team settings