Workshop 3: Chat & Conversations Mastery

Duration: 45 minutes | Level: Intermediate | Prerequisites: Workshops 1-2

What You’ll Master

Take your expert interactions to the next level by mastering every chat feature the B-Bot platform offers.

Chat Interface Deep Dive

Explore every button, menu, and feature in the chat interface

Voice Features

Use voice input and text-to-speech for natural conversations

File Handling

Upload and analyze documents, images, and data files

Memory System

Understand how your expert remembers and uses information

Advanced Features

QA marking, private sessions, and HITL mode

The Chat Interface

When you select an expert and navigate to Chat, you’ll see a rich interface designed for powerful conversations.

Welcome Screen

When starting a new chat, you’ll see:

Expert Greeting: “Hello I am [Expert Name]” with their profile picture
Helper Text: “How can I help you?”
Template Buttons: Quick-start prompts specific to that expert (e.g., “Example template 1 for Alice Johnson”)

Chat Mode Toggle

At the top of the chat area, you’ll see:

Chat Mode / Workspace OFF/ON - Toggle between standard chat and DeepAgents workspace mode

Message Input Area

The input area at the bottom includes:

Text Input: “Type a message…” placeholder
Send Button: Submit your message
Toolbar: Access to all chat features

The toolbar provides quick access to all chat features:

Button	Feature	Description
menu	Menu	Toggle toolbar visibility
Choose Abilities	Abilities	View and activate expert behaviors
Choose previous conversations	History	Browse and resume past chats
model_training	Model	Change the AI model being used
Change standard settings	Settings	Configure temperature, tokens, etc.
Show Memories	Memories	View and manage expert’s memory
Enable/Disable connected Apps	Apps	Toggle integrations and tools
Show/Hide Tasks	Tasks	View DeepAgent task panel
Start new chat	New Chat	Clear and start fresh conversation
Add files to chat	Upload	Add documents, images, or data
Toggle Voice Input	Voice	Enable voice recording mode

Working with Abilities

Viewing Available Abilities

Click the Choose Abilities button in the toolbar to open the abilities panel:

Abilities Panel Structure

Element	Description
Help button	Opens contextual help about abilities
”Select an Ability” heading	Panel title
”Select an ability to enhance your prompt”	Helper text explaining the purpose
Abilities list	Scrollable list of all configured abilities
Close button	Dismiss the panel

Each Ability Card Shows

Ability name - The title of the ability
“Use Ability” button - Click to activate this behavior

Activating an Ability

Open Abilities Panel

Click the Choose Abilities button in the toolbar

Select Ability

Click the Use Ability button on your desired behavior

Observe Change

The expert will now follow that behavioral mode for subsequent messages

Abilities act as “behavioral modes” that shape how your expert responds. You can switch between them anytime during a conversation!

🎯 Exercise: Ability Switching

Start a conversation with your TechSupport AI
Ask a general question
Activate “Troubleshooting Mode” ability
Ask the same question again
Notice how the response style changes!

Chat History

Browsing Past Conversations

Your conversation history shows:

Thread Name: Auto-generated or custom title
Last Message: Preview of recent activity
Date: When the conversation occurred

Managing Conversations

Resume

Click any conversation to continue where you left off

Search

Use the search bar to find specific conversations

Delete

Remove old conversations you no longer need

Model Selection

Switching Models Mid-Conversation

You can switch models during a conversation! The new model will have access to the full conversation history.

When to Switch Models

Scenario	Recommended Model
Quick questions	GPT-4o-mini
Complex analysis	GPT-4o or Claude 3.5
Creative writing	Claude 3.5 Sonnet
Code assistance	GPT-4o
Cost-sensitive	GPT-4o-mini or Gemini

Chat Settings

Click the Change standard settings button in the toolbar to open the Settings dialog.

Settings Dialog Structure

The Settings dialog has a clean, organized layout:

Element	Description
Help button	Opens contextual help about settings
”Settings” heading	Dialog title
”Change settings to get better results”	Helper text

Available Settings

Temperature Slider

Interactive slider that controls randomness in responses:

0.0-0.3: Very consistent, deterministic
0.4-0.7: Balanced creativity and consistency
0.8-1.0: More creative, varied responses

The current value is displayed next to the label (e.g., “Temperature: 0.1”)

Dialog Buttons

Button	Action
Close	Dismiss dialog without saving changes
Save	Apply changes to the current session

Voice Features 🎤

Voice Input

B-Bot supports voice input for hands-free interaction:

Enable Voice Mode

Click the microphone icon to toggle voice mode

Hold to Record

Press and hold the microphone button while speaking

Release to Send

Release when done - your speech is transcribed and sent automatically

Voice Input Tips

✅ Best Practices

Speak clearly at normal pace
Use a quality microphone
Minimize background noise
Pause briefly between sentences

❌ Avoid

Speaking too fast
Background music/TV
Very long recordings
Mumbling or unclear speech

Audio Output 🔊

If text-to-speech is enabled on your expert:

Feature	Description
Auto-play	Responses play automatically
Manual Play	Click play button to hear response
Speed Control	Adjust playback speed
Voice Selection	Choose from various TTS voices

Supported TTS Providers

OpenAI TTS

Voices: Alloy, Echo, Fable, Onyx, Nova, Shimmer

ElevenLabs

Premium natural voices with emotion

Google TTS

WaveNet voices, many languages

Azure Speech

Neural voices, enterprise-grade

File Handling 📎

Uploading Files

Click the upload button or drag files into the chat:

Documents

PDF, Word, Text files

Images

PNG, JPG, GIF, WebP

Data

CSV, JSON, Excel files

🎯 Exercise: File Analysis

Find a PDF document (user manual, report, etc.)
Upload it to your chat
Ask: “Summarize the main points of this document”
Ask follow-up questions about specific sections

Multimodal Conversations

You can combine different input types:

[Text] "What's in this image and how does it relate to the spreadsheet?"
[Image] product_photo.png
[File] inventory.csv

Memory System 🧠

How Memory Works

Your expert can remember information across conversations:

┌─────────────────────────────────────────────────┐
│                  MEMORY SYSTEM                   │
├─────────────────────────────────────────────────┤
│                                                 │
│   User Input ───► Memory Check ───► Response    │
│        │              │              │          │
│        ▼              ▼              ▼          │
│   "Remember       "User prefers    Personalized │
│    I prefer        detailed        response     │
│    detailed        explanations"   with extra   │
│    explanations"                   detail       │
│                                                 │
└─────────────────────────────────────────────────┘

What Gets Remembered

User Preferences

Communication style, detail level, tone preferences

Important Facts

Names, roles, project details, key information

Context

Ongoing projects, previous decisions, related topics

Instructions

Custom rules and behaviors you’ve specified

Managing Memories

Click the Show Memories button in the toolbar to open the Memories panel: Memories Panel Structure:

Element	Description
Help button	Opens contextual help about memories
”Memories” heading	Panel title
”inspect the memory and knowledge containers of your expert”	Helper text
Close button	Dismiss the panel

Memory Management Actions:

View: See all stored memories
Edit: Modify incorrect information
Delete: Remove outdated memories
Add: Manually add important facts

🎯 Exercise: Memory Training

Tell your expert: “Remember that I prefer bullet points over paragraphs”
Ask a question that requires a detailed answer
Check if the response uses bullet points
View the memories panel to confirm it was saved

QA Marking for Training

QA marking lets you create training data from real conversations:

How to Mark QA Pairs

Enable QA Mode

Click the three dots on a message and select “Start QA Marking”

Select Question

Click the user message you want as the question

Select Answer

Click the expert response you want as the answer

Confirm

Review and save the QA pair for training

QA pairs are used to fine-tune your expert and improve its responses over time!

Connectors & Tools

Click the Enable/Disable connected Apps button in the toolbar to open the Connectors & Tools panel.

Connectors & Tools Panel Structure

Element	Description
Help button	Opens contextual help about tools
”Connectors & Tools” heading	Panel title
”Control the Apps you connected or internally available”	Helper text

Tool List

Each tool in the panel displays:

Checkbox - Toggle to enable/disable the tool for this conversation
Tool Name (heading) - e.g., “MEMORY”, “KNOWLEDGE TEST5”
Description - Brief explanation of what the tool does
Tool Image - Visual icon representing the tool

Available Tool Types

MEMORY

Access the expert’s memory system for personalized responses

KNOWLEDGE

Query knowledge bases you’ve created for this expert

Web Search

Search the web for real-time information

API Connectors

External service integrations

Managing Active Tools

In the Tools panel, you can:

Enable/Disable individual tools with checkboxes
View tool descriptions and capabilities
Monitor tool usage in conversation

Disabling unnecessary tools can reduce token usage and improve response speed!

Tool Events in Chat

When your expert uses tools, you’ll see visual feedback:

Web Search

Search result cards with titles, snippets, relevance

API Calls

Clean display of requests and responses

File Operations

Visual feedback for reads/writes

Calculations

Math and data processing results

Advanced: Deep Mode & HITL

DeepAgents Mode

Enable DeepAgents for a persistent workspace:

Tasks

Create and track todos

Files

Read, write, manage files

HITL (Human-in-the-Loop)

When enabled, the agent asks for approval before actions: Auto Mode:

Agent: ✓ Creating task...
Agent: ✓ Complete!

HITL Mode:

Agent: ❓ I propose to create 3 tasks. Approve?
User: [Approve] [Reject] [Modify]

🎯 Challenge Exercise

Complete these advanced chat tasks:

Voice Conversation

Have a full conversation using only voice input

File Analysis

Upload an image and a related document, ask for comparison

Memory Test

Create 3 memories and verify they persist across sessions

QA Creation

Mark 5 good QA pairs from your conversations

Key Takeaways

Rich Interface

The chat interface has many powerful features - explore them all!

Multimodal

Voice, text, files, and images can all be combined

Memory

Teach your expert to remember important information

QA Training

Real conversations become training data

Next Steps

Workshop 4: Training

Learn to train and fine-tune your experts

Chat Documentation

Detailed chat feature reference

Getting Started

Experts

Chat

Training & Knowledge

Apps & Integrations

DeepAgents

Distribution

Account & Settings

Advanced Features

Declarations

​Workshop 3: Chat & Conversations Mastery

​What You’ll Master

​The Chat Interface

​Welcome Screen

​Chat Mode Toggle

​Message Input Area

​Toolbar Buttons

​Working with Abilities

​Viewing Available Abilities

​Abilities Panel Structure

​Each Ability Card Shows

​Activating an Ability

​🎯 Exercise: Ability Switching

​Chat History

​Browsing Past Conversations

​Managing Conversations

Resume

Search

Delete

​Model Selection

​Switching Models Mid-Conversation

​When to Switch Models

​Chat Settings

​Settings Dialog Structure

​Available Settings

​Dialog Buttons

​Voice Features 🎤

​Voice Input

​Voice Input Tips

✅ Best Practices

❌ Avoid

​Audio Output 🔊

​Supported TTS Providers

OpenAI TTS

ElevenLabs

Google TTS

Azure Speech

​File Handling 📎

​Uploading Files

Documents

Images

Data

​🎯 Exercise: File Analysis

​Multimodal Conversations

​Memory System 🧠

​How Memory Works

​What Gets Remembered

User Preferences

Important Facts

Context

Instructions

​Managing Memories

​🎯 Exercise: Memory Training

​QA Marking for Training

​How to Mark QA Pairs

​Connectors & Tools

​Connectors & Tools Panel Structure

​Tool List

​Available Tool Types

MEMORY

KNOWLEDGE

Web Search

API Connectors

​Managing Active Tools

​Tool Events in Chat

Web Search

API Calls

File Operations

Calculations

​Advanced: Deep Mode & HITL

Workshop 3: Chat & Conversations Mastery

What You’ll Master

The Chat Interface

Welcome Screen

Chat Mode Toggle

Message Input Area

Toolbar Buttons

Working with Abilities

Viewing Available Abilities

Abilities Panel Structure

Each Ability Card Shows

Activating an Ability

🎯 Exercise: Ability Switching

Chat History

Browsing Past Conversations

Managing Conversations

Model Selection

Switching Models Mid-Conversation

When to Switch Models

Chat Settings

Settings Dialog Structure

Available Settings

Dialog Buttons

Voice Features 🎤

Voice Input

Voice Input Tips

Audio Output 🔊

Supported TTS Providers

File Handling 📎

Uploading Files

🎯 Exercise: File Analysis

Multimodal Conversations

Memory System 🧠

How Memory Works

What Gets Remembered

Managing Memories

🎯 Exercise: Memory Training

QA Marking for Training

How to Mark QA Pairs

Connectors & Tools

Connectors & Tools Panel Structure

Tool List

Available Tool Types

Managing Active Tools

Tool Events in Chat

Advanced: Deep Mode & HITL

DeepAgents Mode

HITL (Human-in-the-Loop)

🎯 Challenge Exercise

Key Takeaways

Next Steps