Skip to content

README.md

Latest commit

 

History

History
222 lines (179 loc) · 7.33 KB

README.md

File metadata and controls

222 lines (179 loc) · 7.33 KB

DaveFelix Coding Engine

Autonomous code generation system that generates complete production-ready projects from JSON requirements using 37+ specialized AI agents.

Requirements JSON  ──>  Epic Orchestrator  ──>  Complete NestJS/React Project
                           |
                    37+ AI Agents (Society of Mind)
                    Kilo CLI + GPT-5.4 / Claude
                    Docker Sandbox + VNC Preview

Quick Start

# 1. Clone
git clone --recursive https://github.com/Flissel/DaveFelix-Coding-Engine.git
cd DaveFelix-Coding-Engine

# 2. Setup (checks prerequisites, prompts for API keys, builds Docker images)
bash setup.sh

# 3. Start
bash start.sh          # Core: PostgreSQL + Redis + API
bash start.sh --all    # Full: + Sandbox, Gitea, VSCode, Worker

# 4. Stop
bash stop.sh           # Stop all
bash stop.sh --clean   # Stop + wipe volumes

Prerequisites

  • Docker Desktop (includes Docker Compose v2)
  • Python 3.10+
  • Node.js 18+
  • Git

Required API Keys

Key Purpose Get it
OPENAI_API_KEY Code generation (GPT-5.4) platform.openai.com
GITHUB_TOKEN Git push, PR management github.com/settings/tokens

Optional: ANTHROPIC_API_KEY, OPENROUTER_API_KEY, DISCORD_BOT_TOKEN (see .env.example)

Architecture

+-----------------------------------------------------------------+
|  LAYER 1: Society of Mind (37+ Agents)                          |
|  EventBus (push) + SharedState + Convergence Loop               |
|  src/agents/ + src/mind/ + src/engine/                          |
+-----------------------------------------------------------------+
        |
+-----------------------------------------------------------------+
|  LAYER 2: Epic Orchestrator + MCP Tools                         |
|  Parallel task pipeline, Kilo CLI, AutoGen 0.4+ teams           |
|  mcp_plugins/servers/grpc_host/                                 |
+-----------------------------------------------------------------+
        |
+-----------------------------------------------------------------+
|  LAYER 3: MCP Plugin Servers (20+)                              |
|  Playwright, Docker, Redis, GitHub, Prisma, Filesystem, etc.    |
|  mcp_plugins/servers/                                           |
+-----------------------------------------------------------------+
        |
+-----------------------------------------------------------------+
|  OUTPUT: Production-ready project                               |
|  TypeScript/NestJS + React + Prisma + Docker + Tests            |
+-----------------------------------------------------------------+

Services & Ports

Service Port URL Description
API Server 8000 http://localhost:8000 FastAPI control plane
PostgreSQL 5432 — Primary database
Redis 6382 — Task queue + caching
Sandbox VNC 6090 http://localhost:6090 Live preview (noVNC)
App Preview 3100 http://localhost:3100 Generated app
Gitea 3000 http://localhost:3000 Git server
VSCode 8444 http://localhost:8444 Code editor (pw: dev123)
Automation UI 8007 http://localhost:8007 Debug agent

Project Structure

DaveFelix-Coding-Engine/
|-- setup.sh / start.sh / stop.sh    # Setup & lifecycle scripts
|-- docker-compose.yml               # All services
|-- .env.example                     # Environment template
|-- requirements.txt                 # Python dependencies
|
|-- src/
|   |-- agents/                      # 37+ autonomous AI agents
|   |-- mind/                        # EventBus, SharedState, Orchestrator
|   |-- engine/                      # HybridPipeline, Slicer, Merger
|   |-- api/                         # FastAPI server (main.py)
|   |-- tools/                       # Claude CLI, sandbox, test runner
|   |-- mcp/                         # MCP Event Bridge
|   |-- validators/                  # Build, TypeScript, NoMock validators
|   +-- autogen/                     # Kilo CLI wrapper, task enricher
|
|-- mcp_plugins/servers/             # 20+ MCP tool servers
|   |-- grpc_host/                   # Epic orchestrator, task executor
|   |-- fungus_mcp/                  # Project context tools for Kilo
|   |-- playwright/ docker/ redis/   # Infrastructure tools
|   +-- github/ prisma/ npm/ ...     # Integration tools
|
|-- config/
|   |-- engine_settings.yml          # Models, timeouts, consolidation
|   |-- llm_models.yml               # LLM provider configuration
|   +-- society_defaults.json        # Agent feature flags
|
|-- infra/docker/                    # Dockerfiles (api, sandbox, worker)
|-- Data/                            # Project specs + generated output
|-- dashboard-app/                   # Electron/React dashboard (optional)
+-- .claude/skills/                  # 12 skill definitions for agents

How It Works

1. Input: Requirements JSON

{
  "name": "whatsapp-messaging-service",
  "type": "nestjs",
  "features": [
    {"id": "phone-registration", "priority": "high"},
    {"id": "real-time-messaging", "priority": "high"}
  ]
}

2. Epic Orchestrator Pipeline

  1. Parse requirements into Epics (EPIC-001 to EPIC-007)
  2. Split each Epic into ~50-60 consolidated tasks (schema, API, frontend, tests)
  3. Execute tasks in parallel via Kilo CLI + GPT-5.4
  4. Verify each task (TypeScript check, build, tests)
  5. Fix failures automatically via SoM agent loop
  6. Deploy to Docker sandbox with VNC preview

3. Output

Complete production-ready project:

  • NestJS backend with all endpoints
  • React frontend with all pages
  • Prisma database schema + migrations
  • JWT auth with RBAC
  • Docker + CI/CD configs
  • Integration tests (no mocks)

Configuration

engine_settings.yml

models:
  generation: gpt-5.4     # Primary code gen
  fixing: gpt-5.4         # Bug fixing
  planning: gpt-5.4-mini  # Task planning

generation:
  max_parallel_tasks: 4
  consolidation_mode: feature  # granular|endpoint|feature
  max_task_retries: 3

LLM Providers

The engine supports multiple LLM backends:

  • OpenAI (GPT-5.4) — primary, via Kilo CLI
  • Anthropic (Claude) — via Agent SDK
  • OpenRouter — free models for secondary tasks
  • Ollama — local models (qwen2.5-coder)

API Endpoints

Endpoint Method Description
/health GET Health check
/api/v1/projects GET/POST Project management
/api/v1/jobs POST Start generation job
/api/v1/dashboard/generation/{id}/pause POST Pause generation
/api/v1/dashboard/generation/{id}/resume POST Resume with feedback
/ws WS Real-time dashboard updates

Development

# Run tests
pytest

# Run specific test
pytest tests/mind/test_push_architecture.py -v

# View API logs
docker compose logs -f api

# View sandbox logs
docker compose logs -f sandbox

# Rebuild API container after code changes
docker compose build api && docker compose up -d api

Connected Projects

Project Purpose
Minibook Agent collaboration (legacy mode)
DaveLovable Vibe coding UI
OpenClaw Master orchestrator gateway

License

MIT