Skip to content

Two-Weeks-Team/vibeDeploy

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

44 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Hackathon Badge

vibeDeploy

One sentence. One meeting. One live app.

Live Demo

License: MIT Issues PRs Python Next.js LangGraph Gradient ADK

WhatCouncilBrainstormHowArchitectureStackStartDeploy


What is vibeDeploy?

vibeDeploy takes a sentence describing your app idea — or a YouTube URL — and turns it into a live, deployed application on DigitalOcean. No coding required.

Between your idea and the deployed app, The Vibe Council — a panel of 6 specialized AI agents — holds a live debate meeting to evaluate feasibility, challenge assumptions, and score your idea before a single line of code is written.

"I want a restaurant queue management app with QR codes"
                          ↓
              🏛️ The Vibe Council convenes
              🏗️ Architect → 🔭 Scout → 🛡️ Guardian → ⚡ Catalyst → 🎯 Advocate
              🧭 Strategist synthesizes → Vibe Score™: 78.5 → 🟢 GO
                          ↓
              📄 PRD + Tech Spec + API Spec generated
              💻 Frontend + Backend code generated
              🚀 Deployed to DigitalOcean
                          ↓
              ✅ https://your-app.ondigitalocean.app

Key Differentiators

Traditional Vibe Coding vibeDeploy
Generates code only Idea → live deployed app
Single AI writes code 6 AI experts debate before building
You judge quality AI team self-validates feasibility
Deployment is separate Auto-deploy + live URL
Text input only YouTube URL also works
One mode fits all Evaluate or Brainstorm — you choose

The Vibe Council

vibeDeploy's unique decision-making framework. Six specialized AI agents engage in structured debate — challenging, questioning, and refining your idea through a live meeting.

Agent Role Focus Core Question
🏗️ Architect Technical Lead Tech stack, complexity, feasibility "How do we build this?"
🔭 Scout Market Analyst Competition, trends, market fit "Who wants this?"
🛡️ Guardian Risk Assessor Security, legal, failure modes "Why could this fail?"
Catalyst Innovation Officer Uniqueness, disruption, impact "What makes this special?"
🎯 Advocate UX Champion User perspective, accessibility "Will people use this?"
🧭 Strategist Session Lead Synthesis, scoring, verdict "What's the verdict?"

4-Phase Meeting

Phase 1: INDIVIDUAL ANALYSIS    → 5 agents analyze in parallel
Phase 2: CROSS-EXAMINATION      → Structured debate (Architect↔Guardian, Scout↔Catalyst, Advocate challenges)
Phase 3: SCORING                 → Each agent scores their axis (0-100)
Phase 4: VERDICT                 → Strategist calculates Vibe Score™ → GO / CONDITIONAL / NO-GO

Vibe Score™

Vibe Score™ = (Tech × 0.25) + (Market × 0.20) + (Innovation × 0.20) + ((100 - Risk) × 0.20) + (UserImpact × 0.15)

≥ 70 → 🟢 GO          Build and deploy immediately
50-69 → 🟡 CONDITIONAL  Propose scope reduction, user decides
< 50  → 🔴 NO-GO        Detailed failure report + alternatives

Brainstorm Mode

Not ready to evaluate? Brainstorm first. The same 5 council agents switch from judge mode to creative ideation mode — generating possibilities instead of scores.

Your idea → 💡 Brainstorm Mode
                    ↓
  🏗️ Architect: "3 innovative tech stack combos..."
  🔭 Scout:     "3 untapped market segments..."
  ⚡ Catalyst:  "3 ways to make this 10x more disruptive..."
  🛡️ Guardian:  "3 risks turned into competitive advantages..."
  🎯 Advocate:  "3 delightful UX micro-interactions..."
                    ↓
  🧭 Strategist synthesizes all insights into:
     • Top 5 ranked ideas across agents
     • Recurring themes & synergies
     • 3 quick wins to implement first
     • Strategic direction recommendation

Each agent generates ideas, opportunities, wild cards, and action items from their unique perspective. The Strategist then synthesizes everything into a coherent creative brief — no scoring, no judgment, pure creative fuel.


How It Works

Input → Meeting → Build → Deploy

┌──────────────────────────────────────────────────────────────┐
│  INPUT                                                        │
│  Text: "Build a recipe sharing community with AI suggestions" │
│  — or —                                                       │
│  YouTube: https://youtube.com/watch?v=...                     │
│  (auto-extracts idea from transcript + key frame analysis)    │
└──────────────────┬───────────────────────────────────────────┘
                   ▼
┌──────────────────────────────────────────────────────────────┐
│  THE VIBE COUNCIL (Live AI Debate)                            │
│                                                               │
│  Phase 1: 5 agents analyze independently (parallel)           │
│  Phase 2: Cross-Examination debates                           │
│  Phase 3: Each agent scores their axis                        │
│  Phase 4: Strategist → Vibe Score™ → Decision                │
└──────────────────┬───────────────────────────────────────────┘
                   ▼
┌──────────────────────────────────────────────────────────────┐
│  BUILD (if GO or CONDITIONAL-accepted)                        │
│                                                               │
│  📄 PRD + Tech Spec + API Spec + DB Schema                   │
│  💻 Next.js frontend + FastAPI/Express backend                │
│  🎨 App logo + placeholder images (AI-generated)              │
└──────────────────┬───────────────────────────────────────────┘
                   ▼
┌──────────────────────────────────────────────────────────────┐
│  DEPLOY                                                       │
│                                                               │
│  📦 GitHub repo created automatically                         │
│  🚀 Deployed to DigitalOcean App Platform                     │
│  🔗 Live URL: https://your-app.ondigitalocean.app            │
└──────────────────────────────────────────────────────────────┘

Architecture

┌─────────────────────────────────┐     ┌──────────────────────────────┐
│  Frontend (App Platform)        │     │  Agent Backend (Gradient ADK) │
│  Next.js 15 + shadcn/ui        │────▶│  Python + LangGraph           │
│  Static Site                    │◀────│  @entrypoint → POST /run      │
│  .do/app.yaml                   │ SSE │  .gradient/agent.yml          │
└─────────────────────────────────┘     └──────────────────────────────┘
                                                     │
                          ┌──────────────────────────┼──────────────────┐
                          ▼                          ▼                  ▼
                   ┌─────────────┐          ┌──────────────┐   ┌──────────────┐
                   │ Gradient AI │          │ Managed      │   │ DO Spaces    │
                   │ Inference   │          │ PostgreSQL   │   │ (S3)         │
                   │ (8 models)  │          │              │   │              │
                   └─────────────┘          └──────────────┘   └──────────────┘

Two independent deployments:

  • Agentgradient agent deploy to ADK infrastructure (FREE during preview)
  • Frontend → DO App Platform via .do/app.yaml

Tech Stack

Core

Component Technology DigitalOcean Service
Frontend Next.js 15 + shadcn/ui + Tailwind CSS App Platform
Agent Backend Python 3.12 + Gradient ADK + LangGraph Gradient ADK
Database PostgreSQL 16 Managed PostgreSQL
Storage S3-compatible object storage Spaces
AI Models 4 open-source model families via Serverless Inference Gradient AI Platform
Knowledge Base RAG over DO docs + framework patterns Gradient KB

DigitalOcean Gradient AI Features Used (8)

# Feature Usage
1 ADK (Agent Development Kit) Agent system hosting
2 Serverless Inference (Text) 6 council agents + code generation
3 Serverless Inference (Image) App logos, UI mockups, placeholders
4 Knowledge Bases (RAG) DO docs, framework best practices
5 Multi-Agent Routing LangGraph Send API parallel execution
6 Evaluations Agent response quality measurement
7 Tracing Full pipeline debugging
8 Function Calling DO API, GitHub API, YouTube API

Model Assignment

Role Model Type Cost (per 1M tokens)
Council (5 agents) openai-gpt-oss-120b Open-Source (120B) $0.10 / $0.70
Strategist deepseek-r1-distill-llama-70b Open-Source (70B) $0.99 / $0.99
Cross-Examination deepseek-r1-distill-llama-70b Open-Source (70B) $0.99 / $0.99
Code Generation openai-gpt-oss-120b Open-Source (120B) $0.10 / $0.70
CI Repair openai-gpt-oss-120b Open-Source (120B) $0.10 / $0.70
Doc Generation alibaba-qwen3-32b Open-Source (32B) $0.25 / $0.55
Input Processing openai-gpt-oss-120b Open-Source (120B) $0.10 / $0.70
Decision Gate deepseek-r1-distill-llama-70b Open-Source (70B) $0.99 / $0.99
Brainstorm Agents openai-gpt-oss-120b Open-Source (120B) $0.10 / $0.70
Brainstorm Synthesis deepseek-r1-distill-llama-70b Open-Source (70B) $0.99 / $0.99
Web Search mistral-nemo-instruct-2407 Open-Source (12B) $0.30 / $0.30
Image Generation fal-ai/flux/schnell Open-Source (fal) ~$0.003 / image

Multi-model open-source AI: 4 model families via DO Serverless Inference Cost per full deployment: ~$0.12 | $200 credits = ~1,600+ deployments


Getting Started

Prerequisites

Local Development

# Clone
git clone https://github.com/Two-Weeks-Team/vibeDeploy.git
cd vibeDeploy

# Agent Backend
cd agent/
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env  # Fill in your keys
gradient agent run --host 0.0.0.0 --port 8080

# Frontend (new terminal)
cd web/
npm install
NEXT_PUBLIC_AGENT_URL=http://localhost:8080 npm run dev

Open http://localhost:3000 and enter your idea.


Deployment

Live at: https://vibedeploy-7tgzk.ondigitalocean.app

Full Stack → App Platform

The FastAPI backend and Next.js frontend are both deployed via DigitalOcean App Platform:

doctl apps create --spec .do/app.yaml
# Or push to main → auto-deploy (deploy_on_push: true)
# → https://vibedeploy-7tgzk.ondigitalocean.app

Gradient AI Agent (API)

# Create agent via doctl (attach Knowledge Base for RAG)
doctl gradient agent create \
  --name "vibedeploy-agent" \
  --project-id "<project-id>" \
  --model-id "<model-uuid>" \
  --region "tor1" \
  --instruction "Your agent instruction" \
  --knowledge-base-id "<kb-uuid>"

Knowledge Base (RAG)

doctl gradient knowledge-base create \
  --name vibedeploy-docs \
  --region tor1 \
  --project-id "<project-id>" \
  --embedding-model-uuid "<embedding-model-uuid>" \
  --data-sources '[{"web_crawler_data_source":{"base_url":"https://docs.digitalocean.com/products/app-platform/","crawling_option":"UNKNOWN","embed_media":false}}]'

Project Structure

vibeDeploy/
├── web/                          # Next.js Frontend → App Platform
│   ├── src/app/                  # Pages (landing, meeting, brainstorm, result)
│   ├── src/components/           # UI components (council, brainstorm, scores, deploy)
│   └── src/lib/                  # SSE client, API helpers
├── agent/                        # Python Agent → Gradient ADK
│   ├── main.py                   # @entrypoint (Gradient ADK)
│   ├── server.py                 # Local FastAPI server (dev)
│   ├── graph.py                  # Evaluation pipeline (LangGraph StateGraph)
│   ├── graph_brainstorm.py       # Brainstorm pipeline (LangGraph StateGraph)
│   ├── nodes/                    # Pipeline nodes (input, council, brainstorm, build)
│   ├── council/                  # 6 Vibe Council agent definitions
│   ├── tools/                    # YouTube (yt-dlp), GitHub, DO, search tools
│   └── .gradient/agent.yml       # ADK config (3 lines)
├── .do/app.yaml                  # App Platform spec (frontend only)
├── docs/reference/               # 10 planning documents
└── LICENSE                       # MIT

Documentation

Comprehensive planning docs in docs/reference/:

# Document Description
01 Hackathon Rules Rules, timeline, $200 credits guide
02 Judging Criteria 4 criteria × 25% each
03 Submission Checklist Devpost requirements
04 Prizes $20K total, 6 categories
05 Gradient AI Platform SDK patterns, models, pricing
06 DO Products App Platform, DB, Spaces
07 Winning Strategy Demo video, judge targeting
08 Product Concept Vibe Council, scoring system
09 Agent Architecture LangGraph, state, nodes, prompts
10 Technical Plan Deployment, costs, DB schema

Hackathon

Event DigitalOcean Gradient AI Hackathon 2026
Prizes $20,000 total (1st: $5,000 + $2,000 DO credits)
Deadline March 18, 2026, 5:00 PM EDT
Team Two-Weeks-Team

Target Prizes

Prize Why vibeDeploy Fits
1st Place ($5,000) Deep Gradient AI integration (8 features), full pipeline
Best AI Agent Persona 6 distinct agents with personalities debating live
Best Program for the People Democratizes app development for non-coders
Great Whale Prize Largest scope: idea → live deployed app

License

MIT - See LICENSE for details.


Built with DigitalOcean Gradient AI for the DO Gradient AI Hackathon 2026

About

One sentence → AI council debate → live deployed app. 6 specialized AI agents evaluate your idea through structured debate, then generate code and deploy to DigitalOcean.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors