Chat

This guide explains how to use the interactive chat interface to communicate with AI agents in Raikoo. Chat provides a conversational interface for working with agents, testing their capabilities, and iterating on agent configurations.

Accessing Chat

Chat is a project-level tool that allows you to interact with configured agents.

To access chat:

Navigate to your organization's dashboard
Select a project from your projects list
From the project navigation, access Chat under Project Tools

The chat interface loads with an empty conversation, ready for you to select an agent and start conversing.

Chat Interface Overview

The chat interface consists of several key areas:

Header - Contains model/agent selector, new chat button, and menu toggle
Message Area - Displays conversation history with user and assistant messages
Input Area - Text input with attachment support and send controls
Drawer - Side panel with Settings, History, and Workspace tabs (toggle via menu button)

Selecting an Agent

Before sending messages, you can optionally select a configured agent:

Click the menu button in the chat header to open the drawer
Navigate to the Settings tab
Under Agent, select from your project's configured agents
The agent's system prompt, tools, and model preferences are automatically loaded

If you don't select an agent, you can still chat using the default model and tool configurations.

Sending Messages

To send a message to the agent:

Type your message in the input field at the bottom of the chat
Press Enter to send, or Ctrl+Enter (Windows/Linux) or Cmd+Enter (Mac) if you've changed the send preference
The agent processes your message and streams a response

You can change the send behavior in the Settings tab:

Send on Enter (default) - Press Enter to send, Shift+Enter for new line
Send on Ctrl/Cmd+Enter - Press Ctrl/Cmd+Enter to send, Enter for new line

While the agent is generating a response, you can click Stop to cancel the generation.

Multimodal Support

Raikoo chat supports rich multimodal input, allowing you to attach various file types to your messages.

Supported File Types

Images - PNG, JPG, GIF, WebP, SVG
Videos - MP4, WebM, MOV, AVI
Documents - PDF files
Audio - MP3, WAV, M4A, AAC

Attaching Files

To attach files to your message:

Click the attachment button in the chat input area
Select files from your computer, or
Drag and drop files directly into the chat input area

Multiple files can be attached to a single message. Thumbnails appear in the input area showing your attachments. Click the X on any thumbnail to remove that attachment.

The agent receives and processes attachments according to its configured capabilities. For example, vision-enabled models can analyze images, while audio models can transcribe speech.

File Attachments from Workspace

In addition to uploading files from your computer, you can attach files directly from the workspace:

Open the Workspace tab in the drawer
Locate the file you want to attach
Drag the file from the workspace file tree into the chat input area
The file is attached and ready to send

This is useful for discussing files the agent has created or modified during the conversation.

Message Editing

You can edit previously sent user messages to correct mistakes or try different phrasing:

Hover over any of your messages in the conversation
Click the edit button that appears
The message expands into a full-width editor
Make your changes while preserving any attachments
Click Regenerate to resend the edited message

Editing a message truncates the conversation at that point and regenerates the agent's response from there.

Regenerating Responses

You can regenerate the agent's response to any message:

Hover over an assistant message
Click the regenerate button
The agent generates a new response using the same input

This is useful for:

Getting alternative responses
Retrying after an error
Testing agent behavior with the same prompt

Tool Activity

When the agent uses tools during conversation, tool calls are displayed with friendly descriptions and status indicators.

Each tool invocation shows:

Tool name - The human-readable name of the tool being used
Description - A summary of what the tool is doing
Status indicator - Shows whether the tool is running, succeeded, or failed
Results (when available) - The output returned by the tool

Tool activity updates in real-time as the agent processes your request, giving you visibility into the agent's reasoning and actions.

Chat History

All chat conversations are automatically saved as you interact with agents.

To access previous conversations:

Open the drawer via the menu button
Navigate to the History tab
You'll see a list of recent chat sessions

Each session shows:

Timestamp - When the conversation started
Preview - A snippet of the conversation
Agent used (if applicable)

Click any session to load it. The conversation, agent settings, and workspace state are restored exactly as they were.

Starting a New Chat

To start a fresh conversation:

Click the New Chat button in the header
The current conversation is saved to history
A blank chat loads, ready for a new conversation
Agent selection resets to none
The workspace overlay clears

Your model selector preference is preserved across new chats.

Workspace Integration

The chat interface integrates with the project workspace, allowing agents to create and modify files during conversations.

Viewing Workspace Changes

To see what files the agent has created or modified:

Open the drawer
Navigate to the Workspace tab
A file tree shows all files in the workspace with change indicators

Files created or modified during the current conversation are highlighted, making it easy to see what the agent has done.

Workspace State Persistence

The workspace state is automatically saved with each chat session:

When you save a chat session, the workspace overlay is preserved
When you load a chat session from history, the workspace is restored to that state
Starting a new chat clears the workspace overlay

This allows you to resume conversations with the exact workspace context that existed at that point.

Advanced Settings

The Settings tab in the drawer provides fine-grained control over the chat configuration:

Model Selection

Choose which AI model to use for the conversation:

Select from available model providers and models
Or choose a model family to let Raikoo select the best available model

Model selection is preserved when saving chat sessions.

System Prompt

Customize the agent's behavior with a system prompt:

The system prompt defines the agent's role, tone, and capabilities
When you select an agent, its configured system prompt is loaded
You can override or customize the prompt for the current session
System prompts are saved with chat sessions

Tool Configuration

Control which tools are available to the agent:

View all configured tools for the selected agent
Add or remove tools for the current session
Configure tool-specific settings

Tool configurations are saved with chat sessions, allowing you to experiment with different tool combinations.

Token Usage

For each assistant response, you can view detailed token usage statistics:

Hover over an assistant message
Click the usage button that appears
A modal displays:
Input tokens consumed
Output tokens generated
Total tokens used
Cached input tokens (if applicable)
Reasoning tokens (if applicable)

This helps you:

Understand cost implications
Monitor token consumption patterns
Optimize prompts for efficiency

Usage data is also available at the organization level in the Usage page.

Conversation Compaction

When conversations become very long, Raikoo can automatically compact them to stay within model context limits:

Older messages are summarized to reduce token count
Recent messages remain intact for context
Tool calls and results are preserved
A notification indicates when compaction occurs

This allows conversations to continue indefinitely without hitting context limits.

Quick Suggestions

When starting a new conversation, the chat interface may display quick suggestions to help you get started:

Click any suggestion to automatically send it as a message
Suggestions are tailored to the selected agent's capabilities

Best Practices

Agent Selection

Select an agent before starting conversations that require specific tools or capabilities
Test agent configurations in chat before deploying them in workflows
Use descriptive agent names to make selection easier

Message Composition

Be specific in your requests to get better responses
Use attachments when discussing visual content or documents
Break complex requests into multiple messages for better results

Tool Usage

Monitor tool activity to understand what the agent is doing
Verify tool results before taking action based on agent responses
Provide feedback when tools produce unexpected results

History Management

Use descriptive first messages to make sessions easier to find in history
Save important conversations by reviewing them in history
Load previous sessions to continue conversations from where you left off

Workspace Organization

Review workspace changes after agent file operations
Clear workspace between unrelated conversations using New Chat
Use workspace files as attachments for follow-up questions

Troubleshooting Common Issues

Agent Not Responding

Symptoms: Message sent but no response appears

Solutions:

Check that you have a model or agent selected
Verify you have sufficient usage quota
Look for error messages in the chat or notifications
Check network connectivity

Tool Calls Failing

Symptoms: Tool activity shows errors or failures

Solutions:

Verify tool configurations are correct (check Settings tab)
Ensure required permissions are granted (e.g., database write access)
Review tool error messages for specific issues
Check that external services (databases, APIs) are accessible

Attachments Not Processing

Symptoms: Attached files don't appear or aren't processed by agent

Solutions:

Verify file type is supported (images, PDFs, videos, audio)
Check file size limits for your organization
Ensure the selected model supports multimodal input
Try re-uploading the file

History Not Loading

Symptoms: Previous chat sessions don't appear or fail to load

Solutions:

Refresh the page to reload history
Verify you're in the correct project
Check that the chat session wasn't deleted
Ensure you have permission to view project history

Next Steps

Now that you understand the chat interface, you can:

Create and configure agents with specific tools and prompts
Test agent capabilities through interactive conversations
Iterate on configurations by trying different settings and observing results
Use workspace features to collaborate with agents on file-based tasks
Review chat history to learn from previous conversations

The chat interface provides a powerful environment for developing, testing, and using AI agents. By mastering its features, you can effectively leverage agents for a wide variety of tasks.