Skip to main content

Features overview

A quick map of what EdgeWeave can do and where to find it. The chat assistant also uses this page to answer "how do I…" questions.

Building workflows​

The graph editor is the main screen. Drag nodes from the Library panel onto the canvas, or press Ctrl+K to open the node palette and search by name. Connect an output port to an input port to pass data along. Hover a port to see its name, type and description; right-click a node to read its documentation, including whether it supports Python export.

The toolbar has Run and Stop, Save and Open for .weave files, Undo (Ctrl+Z) / Redo (Ctrl+Y), copy and paste, zoom controls and a canvas lock. Each open workflow is a tab; New Module adds one.

After a run, nodes show previews of their results, and the Dataset Inspector lists each node's output type. A node that fails shows a red error preview with the message instead of stopping the whole run.

Sticky notes annotate a graph; they export as comments.

Subsystems​

A .weave file can be called from another graph as a subsystem: its Subsystem Argument nodes become the call's inputs and its Subsystem Return nodes its outputs. Select part of a graph and use Extract selection to Subsystem to move it into its own file. See Subsystems and the subsystem/ examples.

Exporting​

  • Export to Python turns the graph into a standalone script. Nodes share their implementation with the export, so the script computes what the previews show. Nodes that can't be exported say so in their docs and in the script.
  • Export Dashboard publishes a graph's charts, tables and KPIs as a dashboard; you choose which panels to include and how large each is, and can expose parameters. See the dashboard tutorial.
  • Credentials never travel with an export. Password, token and API-key fields are read from environment variables in Python exports (PostgreSQL: PGPASSWORD) and blanked in dashboard exports, and .env files are not copied.

The code editor​

The code editor screen edits project files (Python, JSON, Markdown, …). Its right-hand panel has tabs:

  • Variables — values from the last run.
  • Search — search across project files.
  • Packages — the project's Python packages (each project can have its own virtual environment).
  • Git — commit changes and view diffs, including a visual diff of two versions of a .weave graph (tutorial).
  • Chat — the AI assistant (see below).

Scheduling and the marketplace​

  • The Scheduler runs workflows at a set time or on a repeating schedule while EdgeWeave is open (tutorial).
  • The Marketplace offers extra node modules to install into EdgeWeave (tutorial).

AI assistant​

The Chat tab talks to OpenAI, Anthropic Claude, Google Gemini, Ollama (local models, no key) or any OpenAI-compatible server — pick a provider and a model. API keys go in a .env file (OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY, …): %APPDATA%\EdgeWeave\.env in the installed app, or the repository root when running from source. The chat panel's "not set up" hint shows the exact path. See the chat and RAG tutorial. Chats are saved on your machine; long chats can be summarised into a new chat. The assistant looks up EdgeWeave's nodes, example workflows and docs to answer questions about building graphs.

AI and RAG nodes​

The AI / LLM and AI / RAG node categories call language models from a graph and answer questions over your own documents (load → split → index → retrieve → prompt → LLM). See the ai/rag_demo.weave example and AI / LLM and RAG nodes.

Writing your own nodes​

Put a Python file in your project's modules/user_blocks/ folder and decorate a function with @node (from sdk.node) to describe its ports, fields and Python export. User blocks reload automatically when the file changes. See the tutorial. A user node with the same registered name as a built-in node replaces it.