Skip to main content

Getting started

This page takes you from installing EdgeWeave to running, building and exporting your first workflow. It takes about ten minutes.

1. Install​

On Windows, download the installer from edgeweave.app/download and run it. It installs for your user only, so no administrator prompt appears. You get a Start-menu entry and the app opens when the installer finishes.

EdgeWeave bundles its own Python, so you don't need to install Python first. The first launch takes a little longer while it prepares your data folder:

WhatWhere (Windows)
Your projects (including the bundled demo_project)%APPDATA%\EdgeWeave\projects
API keys for the AI features (.env)%APPDATA%\EdgeWeave\.env
Installed marketplace modules%APPDATA%\EdgeWeave\marketplace_modules

These live outside the install folder, so updating or reinstalling EdgeWeave keeps them.

2. Find your way around​

The tabs along the top switch between screens:

  • Graph Design: the workflow editor, where you'll spend most of your time.
  • Docs: these guides, inside the app.
  • Code Editor: edit project files, inspect variables, manage packages, use Git, and chat with the AI assistant.
  • Settings, Scheduler (run workflows on a timer), Marketplace (extra node modules) and About.

On the Graph Design screen:

  • Library (left): every node, grouped by category. The Files tab next to it browses the project folder.
  • Canvas (middle): your workflow. Scroll to zoom and drag the background to pan.
  • Inspector (right): the output type of each node after a run, plus a history panel.
  • Log (bottom): run messages and errors.
  • Toolbar (top): project picker, New Project, Clear, Open, Save, Copy, Paste, Undo, Redo, Extract selection to Subsystem, Export to Python, Export Dashboard, New Module, and Run / Stop.

3. Run a demo​

  1. Click Open and choose sklearn/sklearn_iris_classification.weave from the demo project.
  2. Click Run (▶).
  3. Each node fills in a preview as it finishes: tables, charts, metrics. If a node fails, it shows a red error preview with the message, and the rest of the run carries on where it can.

The demo project has dozens of examples, one folder per topic: maps/, sqlite/, simweave/, torch/, ai/, dashboard/, subsystem/ and more. Opening a few is the quickest way to see what the nodes can do.

4. Build your own​

  1. Click New Module to open an empty tab.
  2. Add nodes. Drag them from the Library, or press Ctrl+K and type part of a name (see the palette tutorial).
  3. Wire them up by dragging from an output port (right side of a node) to an input port (left side).
  4. Edit a node's fields directly on the node. When a port has the same name as a field and you connect it, the wired value wins and the field greys out.
  5. Hover a port to see its name, type and description. Right-click a node to open its documentation, which also says whether it can be exported to Python.
  6. Click Run, then Save. Workflows are .weave files, which are plain JSON and work well in Git.

A small first graph: CSV Read → DataFrame Describe → Scatter Plot. Point CSV Read at a file in sample_data/.

5. Export it​

  • Export to Python turns the current tab into a standalone script that computes what the previews show. Copy it, download it, or save it. Nodes that can't be exported say so at the top of the script.
  • Export Dashboard turns the charts, tables and KPI tiles into a small web dashboard; see the dashboard tutorial.

Passwords and API keys never end up in an export: Python scripts read them from environment variables, and dashboard exports blank them out.

6. Optional: AI features​

The Chat assistant and the AI / LLM nodes can use OpenAI, Anthropic Claude, Google Gemini, any OpenAI-compatible server, or Ollama (local models, no key needed). For a cloud provider, put its key in the .env file from the table above, for example:

OPENAI_API_KEY=sk-...

EdgeWeave rereads the file before each chat message, so you don't need to restart. The chat panel's provider list shows which providers are set up, and its hint shows the exact path of the file. See the chat and RAG tutorial.

Next steps​

  • Tutorials: the palette, Git and graph diffs, dashboards, chat and RAG, and writing your own node.
  • Features overview: a map of everything else.
  • Subsystems: call one .weave from another.