diagrams-for-ai 0.1.0

AI-friendly architecture diagrams with grid-based positioning

diagrams-for-ai

AI-friendly architecture diagrams with grid-based positioning.

diagrams-for-ai is a Python library that renders beautiful cloud architecture diagrams without Graphviz. Nodes are placed on a simple grid using row and col coordinates, making it trivial for an AI to generate diagrams programmatically.

Why diagrams-for-ai?

• Grid-based positioning -- Place nodes with row and col. No fighting with Graphviz layout.
• AI-friendly -- Explicit coordinates are easy for language models to produce and reason about.
• Real icons -- Leverages the diagrams package for hundreds of icons across AWS, GCP, Azure, Kubernetes, and more.
• Beautiful connections -- Four line styles: curved (bezier), straight, orthogonal, and step.
• SVG & PNG output -- Portable SVG with embedded icons, or raster PNG via Pillow.
• Familiar syntax -- Use >>, <<, and - operators to connect nodes, just like the original diagrams library.
• Mermaid import -- Parse annotated Mermaid flowcharts (.mmd files) into diagrams. Grid positions and icons are embedded as Mermaid comments, so the same file renders in GitHub, IDEs, and diagrams-for-ai.

Installation

uv add diagrams-for-ai

Or with pip:

pip install diagrams-for-ai

The diagrams package (which provides the icons) is installed automatically as a dependency.

Quick example

from diagrams_for_ai import Diagram, Node

with Diagram("Hello", rows=2, cols=2, outformat="png", show=False):
    a = Node("Service A", icon="aws/compute/ec2", row=0, col=0)
    b = Node("Service B", icon="aws/database/rds", row=1, col=1)
    a >> b

Full example

from diagrams_for_ai import Cluster, Diagram, Edge, LineStyle, Node

with Diagram("AWS Web Service", rows=5, cols=7, outformat=["svg", "png"], show=False):

    with Cluster("Public Subnet", row=0, col=1, width=5, height=2,
                 bg_color="#FFF3E0", border_color="#FFB74D"):
        users = Node("Users", icon="aws/general/users", row=0, col=3)
        cdn = Node("CloudFront", icon="aws/network/cloudfront", row=1, col=3)

    with Cluster("Private Subnet", row=2, col=0, width=7, height=3,
                 bg_color="#E8F5E9", border_color="#81C784"):
        lb = Node("ALB", icon="aws/network/elastic-load-balancing", row=2, col=3)
        web1 = Node("Web 1", icon="aws/compute/ec2", row=3, col=1)
        web2 = Node("Web 2", icon="aws/compute/ec2", row=3, col=5)
        db = Node("Aurora", icon="aws/database/aurora", row=4, col=3)

    users >> cdn >> lb
    lb >> [web1, web2]
    web1 >> db
    web2 >> Edge(style="dashed", line_style=LineStyle.ORTHO) >> db

Features

Connection styles

Style	Usage	Description
Curved	a >> b	Smooth bezier curves (default)
Straight	a >> Edge(line_style=LineStyle.STRAIGHT) >> b	Direct lines
Orthogonal	a >> Edge(line_style=LineStyle.ORTHO) >> b	Right-angle routing
Step	a >> Edge(line_style=LineStyle.STEP) >> b	Horizontal-then-vertical

Edge customization

a >> Edge(label="HTTPS", color="#2ECC71", style="dashed") >> b

Clusters

with Cluster("VPC", row=0, col=0, width=4, height=3,
             bg_color="#E8F4FD", border_color="#B0C4DE"):
    # nodes inside the cluster
    ...

Mermaid import

Write your architecture in Mermaid with position annotations in comments:

%% @config name="Hello" rows=2 cols=2
graph TD
    %% @node A pos=0,0 icon=aws/compute/ec2
    %% @node B pos=1,1 icon=aws/database/rds
    A[Service A] --> B[Service B]

Then render it with diagrams-for-ai:

from diagrams_for_ai import Diagram

d = Diagram.from_mermaid_file("architecture.mmd", show=False)
d.render()

See Mermaid Import docs for the full annotation syntax.

Icon providers

AWS, GCP, Azure, Kubernetes, on-premises, and 13 more providers -- all from the diagrams package.

Node("Server", icon="aws/compute/ec2", row=0, col=0)
Node("Pod", icon="k8s/compute/pod", row=0, col=1)
Node("Postgres", icon="onprem/database/postgresql", row=0, col=2)

Development

# Install dependencies
uv sync

# Run tests
uv run pytest

# Serve docs locally
uv run mkdocs serve

License

MIT -- see LICENSE for details.