Vritti

Creative OS MCP — the agent becomes the designer.

A lightweight, no-API-key Creative OS that makes any AI agent a competent graphic designer with brand DNA memory, genre-aware generation, self-critique, and version control.

Features

• Zero API keys — Pollinations.ai (free, no signup) + SVG fallback
• Brand DNA — lock colors/fonts/style once → generate forever
• Genre skills — horror, music, corporate, cinema, minimal, vibrant
• 35+ presets — Instagram, LinkedIn, YouTube, TikTok, album covers, movie posters
• Self-critique — 6-dimension rubric scoring with auto-regenerate
• Version control — every generation tracked (adapted from Rudra Musical Git)
• SVG generation — skill-aware local fallback with rsvg-convert
• Multi-engine — Pollinations → HuggingFace → SVG fallback chain

Install

# No dependencies beyond httpx + pillow (already installed on Termux)
cd ~/vritti
pip install -e .

# Or just use directly
PYTHONPATH=src python3 -m vritti.server

Usage

As MCP Server (stdio JSON-RPC)

# Add to your MCP config
{
  "vritti": {
    "command": "python3",
    "args": ["-m", "vritti.server"],
    "env": { "PYTHONPATH": "/path/to/vritti/src" }
  }
}

Python API

from vritti.server import generate_image, create_brand, quality_check

# Create brand DNA
create_brand(
    name="My Brand",
    primary_color="#1a1a2e",
    secondary_color="#e94560",
    style="dark cinematic",
)

# Generate with brand
result = generate_image(
    prompt="Ancient temple at midnight, dramatic lighting",
    genre="horror",
    brand="My Brand",
    preset="movie-poster",
)

# Quality check
quality = quality_check(
    image_path=result["path"],
    prompt="Ancient temple at midnight",
    genre="horror",
    brand="My Brand",
)
print(f"Score: {quality['overall_score']}/100")

MCP Tools

Tool	Description
generate_image	Generate image from text prompt (Pollinations/SVG fallback)
generate_carousel	Multi-slide carousel for social platforms
create_brand	Create brand DNA profile (colors, fonts, style)
list_skills	List available genres, formats, principles
list_engines	List generation engines and status
list_presets	List 35+ platform presets
list_brands	List saved brand profiles
get_version	Get Vritti version info
get_platform	Get detected platform and output paths
version_history	Get generation version history
quality_check	Evaluate image quality against rubric

Skills

Genres

• horror — deep reds, blacks, monochromatic
• music — genre-dependent (hip-hop bold, lo-fi muted)
• corporate — blues, whites, conservative
• cinema — dramatic, high contrast, color grading
• minimal — whitespace, single accent, breathing room
• vibrant — neon, high saturation, bold gradients

Formats

• instagram-square — 1080x1080
• instagram-carousel — 1080x1350 (4:5)
• album-cover — 3000x3000 (Spotify/Apple requirement)
• movie-poster — 2700x4000 (27:40 at 300 DPI)
• linkedin-carousel — 1080x1350
• youtube-thumbnail — 1280x720

Architecture

vritti/
├── src/vritti/
│   ├── server.py          # MCP server (JSON-RPC over stdio)
│   ├── engines/           # Pollinations + HuggingFace + SVG
│   ├── svg/               # SVG generation + PNG conversion
│   ├── skills/            # Genre/format/principle loaders
│   ├── brand/             # Brand DNA CRUD
│   ├── quality/           # 6-dimension rubric scoring
│   ├── versioning/        # Version control (Rudra-adapted)
│   ├── output/            # 35+ platform presets
│   └── platform.py        # Termux/Android/Windows detection
├── skills/                # JSON skill files
│   ├── genres/            # Horror, music, corporate, etc.
│   ├── formats/           # Instagram, album, poster specs
│   └── principles/        # Color theory, typography, composition
└── pyproject.toml

License

MIT — LGCY Studios