Skip to content

Blueprint

Blueprint is a PyCharm plugin that helps developers plan their codebase before implementing it, making AI-assisted coding more structured, predictable, and easy to reason about.

Core trick

Parses Python classes into editable Mermaid UML, then generates dependency-ordered code nodes that are diffed and applied after review.

Tags

Description

Blueprint is a PyCharm plugin that helps developers plan their codebase before implementing it, making AI-assisted coding more structured, predictable, and easy to reason about. It gives developers a clear architectural overview by analyzing an existing Python project and turning it into editable Mermaid UML, or by letting them sketch a design first and refine it through an AI-assisted side chat. Once the architecture is ready, Blueprint converts the design into scoped, dependency-aware implementation nodes, then guides each change through planning, code generation, review, diff preview, validation, and apply. This creates a safer loop from architecture to code, helping developers use AI with more control, better context, and fewer unreviewed changes.
NoahLundSyrdal/blueprint★ 01 forksKotlinMITCreated Apr 18, 2026 · last push Apr 19, 2026

README excerpt

# Blueprint

Blueprint is a PyCharm/JetBrains plugin for architecture-first Python development.

The core loop is:

1. Open an existing Python project.
2. Abstract the codebase up into editable Mermaid UML.
3. Edit the UML directly or use the side chat to refine the architecture.
4. When the UML is ready, generate scoped code nodes from it.
5. Plan, execute, review, preview diff, and apply approved changes.
6. Repeat the loop whenever needed: codebase -> UML -> chat refinement -> code.

## Current Scope

Blueprint is intentionally focused on a hackathon-ready vertical slice:

- Python project context detection.
- Editable Mermaid UML generation from Python classes.
- OpenAI-assisted UML refinement in the side chat.
- UML conversion into dependency-aware code-generation nodes.
- Offline mock mode for deterministic demos.
- Plan, execute, review, diff preview, and scoped apply.
- List-based node workflow with a read-only mini graph.

It does not yet include drag-and-drop graph editing or true parallel execution.

## Run Locally

Use JDK 17:

```bash
JAVA_HOME=/opt/homebrew/opt/openjdk@17/libexec/openjdk.jdk/Contents/Home ./gradlew runIde
```

For the easiest demo test, use the helper script:

```bash
./scripts/run-blueprint-pycharm.sh
```

That launches a PyCharm sandbox with Blueprint installed and opens:

```text
examples/invite_project
```

Once PyCharm opens, use `View -> Tool Windows -> Blueprint` if the Blueprint tab is not already visible.

Build:

```bash
JAVA_HOME=/opt

Team