Back to Documentation

Using Chat

19 steps - 7 minutes

Learn how to use the Uru chat interface to interact with AI and access your tools.

Chat Features

Uru chat uses Server-Sent Events (SSE) for streaming responses with resume support. Files persist across turns and are included in subsequent AI payloads. Thinking bubble shows human-friendly statuses (processing request, executing tools, reading). Artifacts render in side panels with full screen expansion. Conversations persist across reloads and continue executing in the background when you navigate away.

1

Navigate to Chat

From the sidebar, click on "Chat" to open the chat interface.

2

Select a Persona (Optional)

Click the persona selector at the top to browse and select an AI persona. The default is "Uru Guru" (no persona). Selected personas shape how the AI responds.

3

Add Context (Optional)

Click the context icon to browse and select context blocks. Selected context is included in the AI conversation for relevant background information.

4

Upload Files Before First Message

You can upload files before sending your first message. Click the attachment icon to upload files (PDF, DOCX, TXT, CSV). Files are automatically processed and included in the conversation.

5

File Persistence Across Turns

Uploaded files persist across conversation turns. They are included in subsequent AI payloads until you remove them. Full file text is included for files under the token limit.

6

Type Your Message

Enter your question or request in the chat input field. Be clear and specific about what you need.

7

Send Your Message

Press Enter or click the send button to submit your message. The AI will begin processing your request via streaming.

8

Thinking Bubble - Processing Request

The thinking bubble appears immediately when you send a message. It shows "processing request" while waiting for the first OpenAI API call.

9

Thinking Bubble - Executing Tools

When the AI executes tools, the thinking bubble shows "executing tools" with the tool name. It hides when the first tool appears or AI starts typing.

10

Thinking Bubble - Reading

After a tool completes and before a subsequent OpenAI call, the thinking bubble shows "reading" at the bottom. It disappears during token streaming and reappears during downtime.

11

Tool Execution in Real-Time

When the AI uses tools (like searching, sending emails, or accessing integrations), you see tool execution cards in real-time as they happen.

12

Streaming Response

The AI response streams in real-time via Server-Sent Events (SSE). You see text, code, artifacts, and tool results in chronological order as they happen.

13

Artifacts - Side Panel

Click on artifact bubbles (reddish magenta clickable bubbles) to view rendered code, documents, charts, or interactive components in a side panel with full screen expansion.

14

Artifact Types

Artifacts support multiple types: code (Monaco editor), HTML (iframe with DOMPurify), SVG, Mermaid diagrams, markdown documents, and React components with curated libraries.

15

Continue the Conversation

Ask follow-up questions or provide additional context. The AI maintains conversation history automatically with turn-based formatting enabled by default.

16

Use Prompts

Click the prompts library icon to browse and select prompts. Click a prompt to insert it into your message for common tasks.

17

Conversation Persistence and Background Execution

Conversations persist across reloads and navigation. Chat runs continue executing server-side when you leave the page and resume on return.

18

Deep-Linking with Conversation IDs

After starting a conversation, the URL updates with the conversation ID (e.g., /c/<id>). Reloads and returning to the chat page restore the same active conversation.

19

Start New Conversation

Click "New Chat" to start a fresh conversation. The URL navigates back to the base chat URL. Previous chats remain accessible in your history.

Best Practices

  • Be specific and clear in your requests
  • Select relevant context before starting complex tasks
  • Use personas to get specialized responses
  • Review tool execution results to verify accuracy
  • Save useful prompts for reuse in future conversations

Next Step

Learn how to build automations to automate your workflows.

Building Automations