Chat Node

The Chat Node is the primary way to interact with AI models. It supports back-and-forth conversations, tool calling, context from connected nodes, and multiple AI providers.

---

Sending Messages

Type your message in the input field at the bottom of the chat node.

• Enter submits your message
• Shift+Enter inserts a new line (also works for code blocks and lists)
• Cmd/Ctrl+Shift+Enter splits into a new branch

You can also click the Send button to submit, or the Split button beside it to send your prompt in a new branched chat node.

---

Model Selection

1. Click the model dropdown (bottom left of the chat input)

2. Select your preferred model from any configured provider

3. RHAI models (Basic, Pro, Advanced) show the underlying model name as a sub-label

---

Tools Dropdown

Use the + button in the chat input to access tools and settings for the current node. This dropdown includes:

• RAG Mode — toggle Retrieval Augmented Generation
• Personas — select or create persona prompts
• MCP Tools — enable MCP server tools for the conversation
• Reasoning Effort — configure reasoning depth (Low, Medium, High, Max) for models that support it

---

RAG Mode

Retrieval Augmented Generation (RAG) selectively retrieves only the most relevant information from connected nodes. This optimizes token usage by filtering content before sending to the AI model.

1. Open the + tools dropdown and toggle RAG mode, or click the scissor icon

2. Icon highlights when active

3. RAG token count appears at top of chat after LLM response

4. Only relevant context from connected nodes is used, not all content

For small context sizes (under 10,000 tokens) RAG mode is not recommended.

You can chat with file nodes immediately after adding them without waiting for RAG processing. RAG mode specifically requires processing to be complete.

---

Personas

Create and save custom system prompts:

1. Open the + tools dropdown and navigate to Personas

2. Click "Create Persona"

3. Set name, color, and prompt

4. Click "Create"

To use: Select from the tools dropdown or the persona selector in the input field

---

MCP Tools

Use Model Context Protocol servers to extend what the AI can do:

1. Configure MCP servers in Settings > MCP Servers

2. Open the + tools dropdown in a chat node

3. Enable specific MCP servers for this conversation

4. The AI will use enabled MCP tools when relevant during the conversation

MCP requires models that support tool calling.

---

Reasoning Effort

For models that support extended thinking (like Claude Opus, Codex):

1. Open the + tools dropdown

2. Navigate to Reasoning Effort

3. Select effort level: Low, Medium, High, or Max

Higher reasoning effort gives the model more time to think through complex problems.

---

Audio Input

1. Click the microphone icon or press Ctrl + M

2. Speak into your microphone

3. Click "Done" or press Esc to stop

4. Edit the transcribed text if needed

5. Press Enter to send

---

Regenerating Responses

To regenerate an AI response, click the regenerate icon underneath any assistant message. This replaces the previous "Retry" button that was at the bottom of the chat.

---

Editing Messages

You can edit any previously sent message. The edited message will be re-sent and the conversation will continue from that point.

---

Token Summary

View token usage for any chat node by checking the node handles:

• Incoming tokens — context flowing in from connected nodes
• Own tokens — content within this node's conversation
• Input tokens — total tokens sent to the model

The taker handle shows a tooltip with incoming token counts on hover.