Skip to main content

Graph Copilot

Graph Copilot is a CodefyUI plugin that adds a chat panel to the editor. You describe what you want in plain language and the AI generates a sequence of graph operations (add nodes, connect ports, set parameters) that are applied atomically — one undo step per AI edit. You can stop mid-stream, retry failed requests, and resume a conversation across sessions.

:::note Availability Graph Copilot is built on two CodefyUI features: the plugin frontend extension API and the unified LLM proxy endpoint (/api/llm/chat). Both are in CodefyUI 1.3.0 and later. If cdui --version reports an older version, run cdui update before installing. :::

Installation​

cdui plugin install treeleaves30760/CodefyUI-Plugin-Graph-Copilot

Then reload the editor (press F5 or close and reopen the tab). The Graph Copilot panel appears as a floating widget in the editor.

Plugin source and issues: github.com/treeleaves30760/CodefyUI-Plugin-Graph-Copilot

Quick start​

  1. Install the plugin (above) and reload the editor.
  2. Click the round Graph Copilot button in the bottom-right corner of the canvas to open the chat panel.
  3. Click the settings (gear) icon, choose a provider, and paste your API key — or, for OpenAI Codex, click Sign in and approve in the tab that opens. Pick a model (use Load list to fetch the provider's models).
  4. Type a request such as Build a small MLP classifier and press Enter.
  5. Watch the nodes appear and wire up on the canvas as the AI streams its plan. Press Ctrl+Z once to undo the whole edit, or keep chatting to refine it.

The provider and key only need to be set once — they persist in your browser. The rest of this page covers each part in detail.

Choosing an LLM provider​

Open the settings icon in the Graph Copilot panel to configure your provider and key.

ProviderNotes
OpenAI APIStandard https://api.openai.com/v1 endpoint. Requires an OpenAI API key. Billed per token.
OpenAI Codex (ChatGPT sign-in)Uses the ChatGPT web session. No separate API key required, but subject to ChatGPT usage quotas and OpenAI's ToS — use of the internal session API for automation is a gray area not officially sanctioned by OpenAI.
OpenRouterAggregates many providers under one key. Set the base URL to https://openrouter.ai/api/v1 and select your preferred model.
Claude APIAnthropic's API, accessed through CodefyUI's proxy which translates the OpenAI-compatible request format. Requires an Anthropic API key.
Custom (OpenAI-compatible)Any server that implements the OpenAI /chat/completions endpoint — for example, a local Ollama instance: http://localhost:11434/v1. Set the base URL and optionally a key.

Key handling​

API keys are stored in localStorage under a namespace private to Graph Copilot and never sent to the CodefyUI backend or any third party — only to the provider you have configured. The local CodefyUI backend (/api/llm/chat) acts as a streaming proxy, forwarding your request to the configured provider and streaming the response back; it does not log or persist keys or message content.

Usage​

Sending a request​

Type your request in the chat input and press Enter (or click Send). Examples:

  • "Add a two-layer MLP with ReLU activations"
  • "Connect the CrossEntropy node to the output of the last Linear"
  • "Set the hidden size on Linear-1 to 512"

The AI returns a plan followed by a list of operations. You can see each operation as a chip label (e.g., "add Linear", "add ReLU", "connect") as they are applied.

Conversation history​

The chat history for the current graph is saved in localStorage. When you reopen the editor or reload the page, Graph Copilot resumes the conversation where you left off.

Stop and retry​

Click Stop during a stream to cancel the in-flight request. The partial response is discarded. Click Retry on any AI message to resend that turn with the same context.

Undoing AI edits​

Every AI edit is a single undo snapshot. Press Ctrl+Z (or Cmd+Z on macOS) once to undo the entire batch of operations from the last AI response.

Tips​

  • Give context about what you are building: "I am building a vision classifier with a ResNet backbone" helps the AI make better choices.
  • If the AI adds a node type that does not exist in your palette, it will be skipped and reported — use cdui plugin install to add the required pack first.
  • Graph Copilot reads the current graph state and the full node palette before each request, so it knows what types are available and what is already on the canvas.

See also​