​

AI Chat & Conversations

Backstack’s chat interface lets you have conversations with AI assistants that can execute tools, remember information about you, and help you accomplish tasks across your organization.

​

Starting a Conversation

​

Creating Your First Chat

1. Navigate to Chat in the sidebar
2. You’ll see a welcome message: “How can I help you today?”
3. Type your message in the input field at the bottom
4. Press Enter to send (or Shift+Enter for a new line)

A new conversation is created automatically when you send your first message.

​

Message Input

The chat input supports:

• Enter - Send your message
• Shift+Enter - Add a new line without sending
• Rich text pasting from other applications
• Long messages with automatic scrolling

​

Conversation Features

​

Real-Time Streaming

Assistant responses appear word-by-word as they’re generated:

1. You send a message
2. A typing indicator appears: “Assistant is typing…”
3. The response streams in real-time
4. The message completes and remains in your history

This gives you immediate feedback and lets you see the assistant’s thought process as it works.

​

Tool Execution

When the assistant needs to perform actions, it executes tools automatically: Tool Call Display:

• Tools appear as expandable boxes with a wrench icon
• Status badges show execution progress:
  • “Executing…” (blue) - Tool is currently running
  • “Success ✓” (green) - Tool completed successfully
  • “Failed ✗” (red) - Tool execution failed
• Execution time is displayed in milliseconds

Viewing Tool Details: Click on any tool call box to expand it and see:

1. Parameters - The inputs passed to the tool (formatted JSON)
2. Result - The tool’s output or response
3. Error Message - If the tool failed, why it failed

Tools execute in real-time as the assistant works through your request, appearing in chronological order within the conversation.

​

Memory and Learning

The assistant can remember information about you across conversations: Memory Callouts appear inline with a brain icon:

• Facts: “Memory: User prefers TypeScript”
• Patterns: “Memory: Pattern: email-validation - Examples: [email protected]”

Memories help the assistant:

• Remember your preferences
• Recall information from previous conversations
• Personalize responses to your needs
• Avoid asking repetitive questions

Deleted memories appear with strikethrough styling.

​

Reasoning Sections

For complex tasks, the assistant may show its reasoning process: Reasoning boxes (lightbulb icon) expand to show:

• Progress - What’s been accomplished so far
• Next Action - What will be done next
• Why - Reasoning for the chosen approach
• Plan Adjustments - Any changes to the original plan
• Need Your Input - Questions or decisions requiring your response

This transparency helps you understand how the assistant is solving your problem.

​

Conversation Routing

The assistant can organize your messages across multiple conversations automatically: Context Switch Indicators show routing decisions:

• “Continuing here” (green) - Staying in current conversation
• “Switched conversation” (blue) - Moved to a different conversation
• “New conversation” (purple) - Created a new conversation

Each indicator explains why the switch occurred, helping you understand how your conversations are organized.

​

Managing Conversations

​

Viewing Conversation History

Access your conversation history:

1. Click the History button in the top-right corner
2. View a list of all your conversations
3. Each conversation shows:
   • Title
   • Number of messages
4. Click any conversation to jump to it in the message stream

​

Deleting Conversations

To remove a conversation:

1. Open the History panel
2. Hover over the conversation (or tap on mobile)
3. Click the trash icon
4. Confirm deletion in the dialog

​

Continuous Message Stream

All your messages appear in one continuous, chronological stream:

• Messages from all conversations are sorted by time
• Conversation boundaries are marked with visual separators
• You can see how conversations relate to each other
• Scroll through your complete history seamlessly

This design helps you track patterns and connections across conversations.

​

Message Types

​

Text Messages

Your Messages:

• Light background highlight
• Labeled “You”
• Appear immediately when sent

Assistant Messages:

• Rendered with full Markdown formatting
• Support for:
  • Bold and italics
  • Code blocks and inline code
  • Lists (bulleted and numbered)
  • Tables
  • Blockquotes
  • Links and images
  • Headings

​

Rich Content

Messages can include:

1. Tool Calls - Executable actions with parameters and results
2. Memory Callouts - Facts and patterns the assistant has learned
3. Reasoning - Thought process and planning steps
4. Context Switches - Conversation routing decisions
5. Policy Violations - Content policy notices (if applicable)

All content types are inline and chronologically ordered, giving you a complete view of how the assistant worked through your request.

​

Best Practices

​

Effective Conversations

Be Specific:

• Provide context about what you’re trying to accomplish
• Include relevant details (file names, error messages, requirements)
• Mention any constraints or preferences

Use Multi-Turn Conversations:

• Break complex tasks into steps
• Ask follow-up questions
• Refine your request based on initial responses

Leverage Tools:

• The assistant can execute tools automatically
• Mention what actions you need (search, file operations, API calls)
• Review tool results to verify they’re what you expected

Trust Memory:

• The assistant remembers your preferences
• You don’t need to repeat context in every conversation
• Correct the assistant if it remembers something incorrectly

​

Conversation Organization

Let the Assistant Route:

• The assistant will organize messages across conversations
• Trust the routing decisions - they’re based on context and relevance
• Check the routing indicators to understand why switches occurred

Use History Effectively:

• Review past conversations to recall decisions
• Delete conversations that are no longer relevant
• Keep conversation titles descriptive (the assistant generates these automatically)

Monitor Tool Execution:

• Expand tool calls to verify parameters
• Check execution times for performance insights
• Report failed tools to your organization admin

​

Mobile Usage

The chat interface is optimized for mobile:

• Touch Targets - All buttons are at least 44x44 pixels
• iOS Safari - Keyboard handling and safe area support
• Responsive Layout - Adapts to screen size
• Momentum Scrolling - Smooth code block scrolling
• No Zoom - 16px input font prevents iOS auto-zoom

Access your conversations seamlessly across desktop and mobile devices.

​

Troubleshooting

​

Assistant Not Responding

Problem: Message sent but no response appears Solutions:

• Check your internet connection
• Refresh the page and try again
• Verify your organization has AI providers configured (Organization → AI Providers)
• Contact your organization admin if the issue persists

​

Tool Execution Failures

Problem: Tool shows “Failed” status Solutions:

• Expand the tool call to see the error message
• Common issues:
  • Missing permissions - Tool needs authorization or API key
  • Invalid parameters - Check if the tool received correct inputs
  • Service unavailable - External service may be down
• Report repeated failures to your organization admin

​

Streaming Interrupted

Problem: Response stops mid-stream Solutions:

• Wait a moment - streaming may resume
• Refresh the page if response is stuck for more than 30 seconds
• Check browser console for errors
• Try sending the message again

​

Memory Issues

Problem: Assistant forgot something or remembers incorrectly Solutions:

• Explicitly correct the assistant: “Actually, I prefer Python, not JavaScript”
• The assistant will update its memory based on your correction
• Deleted memories are marked with strikethrough
• Contact support if memory issues persist across sessions

​

Can’t Find Past Conversations

Problem: Conversation missing from history Solutions:

• Scroll through the continuous message stream - it may be there
• Check if you accidentally deleted it
• Verify you’re in the correct workspace (conversations are workspace-specific)
• Try refreshing the page

​

Privacy and Security

​

Data Handling

• Conversations are stored within your workspace
• Only workspace members can see conversations
• Messages are transmitted securely
• Tool executions are logged for audit purposes

​

Content Policies

The assistant follows content policies defined by:

• Your organization’s security policies (see Security Policies)
• Platform-wide safety guidelines
• Tool-specific restrictions

Policy violation notices appear as footnotes on affected messages.

​

Next Steps