[GH-ISSUE #19] [Proposal] Intelligent Project Onboarding & Pattern Analysis #5

New issue

Closed

opened 2026-02-27 10:21:44 +03:00 by kerem · 1 comment

kerem commented

2026-02-27 10:21:44 +03:00

Owner

Originally created by @krzemienski on GitHub (Jan 2, 2026).
Original GitHub issue: https://github.com/mikeyobrien/ralph-orchestrator/issues/19

Feature Proposal: Intelligent Project Onboarding

The Problem

When you first run Ralph on a new project, it doesn't know:

What kind of project this is
How to compile, test, and execute it
What patterns the codebase follows
What tools and conventions are in use

Currently, you have to manually configure or hope Ralph figures it out.

Proposed Solution

An intelligent onboarding system that:

Analyzes the project - Scans structure, config files, dependencies
Detects patterns - Identifies frameworks, conventions, architecture
Generates configuration - Creates a tailored ralph.yml for this project
Learns preferences - Remembers project-specific context for future sessions

Example Flow

$ ralph onboard

🔍 Analyzing project structure...

📋 Project Analysis:

**Detected:**
- Framework: Next.js 14 with App Router
- Language: TypeScript (strict mode)
- Testing: Jest + Playwright
- Styling: Tailwind CSS + shadcn/ui
- State: Zustand
- API: tRPC with Drizzle ORM

**Commands:**
- Development: `npm run dev`
- Production: `npm run compile`
- Tests: `npm test` / `npm run e2e`

**Architecture Patterns:**
- Feature-based directory structure
- Server components by default
- API routes in `/app/api/`
- Shared utils in `/lib/`

**Generating ralph.yml...**

Would you like me to create this configuration? [Y/n]: _

Generated Configuration

# ralph.yml - auto-generated by Ralph onboarding
project:
  name: my-nextjs-app
  type: web
  framework: nextjs

commands:
  compile: npm run compile
  test: npm test

patterns:
  architecture: feature-based
  components: server-first
  api: trpc

context:
  # Ralph will remember these for future sessions
  key_directories:
    - src/features/
    - src/lib/
    - src/app/api/
  conventions:
    - "Use server components by default"
    - "Colocate tests with components"

Benefits

Faster starts: No manual configuration needed
Better results: Ralph understands your project's patterns
Consistency: Follows your existing conventions
Learning: Improves over multiple sessions

Implementation via Self-Improvement

This would be created using Ralph's self-improvement system:

Write a prompt describing the onboarding feature
Run python scripts/self_improve.py -P prompts/ONBOARDING_PROMPT.md
Ralph creates onboarding into itself

Community Input Requested

Detection scope: What project aspects should onboarding detect?
Configuration format: YAML? JSON? Something else?
Interactive vs automatic: How much should be automatic vs user-confirmed?
Memory: Should Ralph persist learned patterns between sessions?

The best tools feel like they already know you. This feature is about making Ralph feel less like a generic AI and more like a colleague who's worked on your codebase before.

What patterns in your projects would you want Ralph to automatically detect? Drop a comment!

Originally created by @krzemienski on GitHub (Jan 2, 2026). Original GitHub issue: https://github.com/mikeyobrien/ralph-orchestrator/issues/19 ## Feature Proposal: Intelligent Project Onboarding ### The Problem When you first run Ralph on a new project, it doesn't know: - What kind of project this is - How to compile, test, and execute it - What patterns the codebase follows - What tools and conventions are in use Currently, you have to manually configure or hope Ralph figures it out. ### Proposed Solution An **intelligent onboarding system** that: 1. **Analyzes the project** - Scans structure, config files, dependencies 2. **Detects patterns** - Identifies frameworks, conventions, architecture 3. **Generates configuration** - Creates a tailored `ralph.yml` for this project 4. **Learns preferences** - Remembers project-specific context for future sessions ### Example Flow ```bash $ ralph onboard 🔍 Analyzing project structure... 📋 Project Analysis: **Detected:** - Framework: Next.js 14 with App Router - Language: TypeScript (strict mode) - Testing: Jest + Playwright - Styling: Tailwind CSS + shadcn/ui - State: Zustand - API: tRPC with Drizzle ORM **Commands:** - Development: `npm run dev` - Production: `npm run compile` - Tests: `npm test` / `npm run e2e` **Architecture Patterns:** - Feature-based directory structure - Server components by default - API routes in `/app/api/` - Shared utils in `/lib/` **Generating ralph.yml...** Would you like me to create this configuration? [Y/n]: _ ``` ### Generated Configuration ```yaml # ralph.yml - auto-generated by Ralph onboarding project: name: my-nextjs-app type: web framework: nextjs commands: compile: npm run compile test: npm test patterns: architecture: feature-based components: server-first api: trpc context: # Ralph will remember these for future sessions key_directories: - src/features/ - src/lib/ - src/app/api/ conventions: - "Use server components by default" - "Colocate tests with components" ``` ### Benefits - **Faster starts**: No manual configuration needed - **Better results**: Ralph understands your project's patterns - **Consistency**: Follows your existing conventions - **Learning**: Improves over multiple sessions ### Implementation via Self-Improvement This would be created using Ralph's self-improvement system: 1. Write a prompt describing the onboarding feature 2. Run `python scripts/self_improve.py -P prompts/ONBOARDING_PROMPT.md` 3. Ralph creates onboarding into itself ### Community Input Requested - **Detection scope**: What project aspects should onboarding detect? - **Configuration format**: YAML? JSON? Something else? - **Interactive vs automatic**: How much should be automatic vs user-confirmed? - **Memory**: Should Ralph persist learned patterns between sessions? --- *The best tools feel like they already know you. This feature is about making Ralph feel less like a generic AI and more like a colleague who's worked on your codebase before.* *What patterns in your projects would you want Ralph to automatically detect? Drop a comment!*

kerem closed this issue

2026-02-27 10:21:44 +03:00

kerem commented

2026-02-27 10:21:44 +03:00

Author

Owner

@krzemienski commented on GitHub (Jan 2, 2026):

Moving this proposal to the fork for now - focusing on validation gates first.

@krzemienski commented on GitHub (Jan 2, 2026): Moving this proposal to the fork for now - focusing on validation gates first.

kerem referenced this issue

2026-02-27 10:22:09 +03:00

[PR #5] [MERGED] feat(acp): Add ACP (Agent Communication Protocol) adapter for CLI agents #82