> ## Documentation Index
> Fetch the complete documentation index at: https://docs.b-bot.space/llms.txt
> Use this file to discover all available pages before exploring further.

# Key Concepts & Basics

> Fundamental concepts of the B-Bot Hub platform

# Key Concepts & Platform Basics

Welcome to B-Bot Hub - a platform for creating, training, and deploying specialized AI experts. This guide covers the fundamental concepts you need to understand to get the most out of the platform.

## Core Architecture

### The Three-Layer Model

B-Bot Hub operates on a simple but powerful three-layer architecture:

```
Layer 1: EXPERTS
    ↓
Layer 2: DISTRIBUTION CHANNELS
    ↓
Layer 3: END USERS
```

<Steps>
  <Step title="Experts (Layer 1)">
    **Your AI Agents**

    Experts are specialized AI agents you create with:

    * Unique personality and identity
    * Custom training and knowledge
    * Specific capabilities and tools
    * Configured modalities (voice, image, etc.)

    Think of experts as your AI employees - each with their own role, expertise, and way of working.
  </Step>

  <Step title="Distribution Channels (Layer 2)">
    **Deployment Methods**

    How your experts reach users:

    * **B-Bot Chat**: Internal chat interface
    * **Embed Widget**: Website chat widgets
    * **API**: Custom integrations
    * **Tasks**: Scheduled automation
    * **Templates**: Public marketplace sharing

    One expert can be deployed through multiple channels simultaneously.
  </Step>

  <Step title="End Users (Layer 3)">
    **Your Audience**

    Who interacts with your experts:

    * Customers
    * Team members
    * Website visitors
    * API consumers
    * Automated systems
  </Step>
</Steps>

## Key Features

### 1. Expert Creation & Management

<CardGroup cols={2}>
  <Card title="8-Step Creation Process" icon="list-ol">
    1. Basic Information
    2. Profile Picture (AI-generated option!)
    3. Abilities & Behaviors
    4. Quick-Start Templates
    5. Models & Modalities
    6. App Connections
    7. Team Configuration
    8. Review & Deploy
  </Card>

  <Card title="Expert Configuration" icon="sliders">
    Each expert includes:

    * Name, profession, description
    * System prompt (personality)
    * Abilities (behavior modes)
    * Templates (quick prompts)
    * Model selection
    * Tools & apps
    * Voice settings
  </Card>
</CardGroup>

### 2. Fine-Tuning & Training

**Transform generic models into specialists:**

<Tabs>
  <Tab title="QA Marking">
    **Collect training data naturally:**

    1. Chat with your expert
    2. When you get a perfect response, mark it
    3. Build up 50-500+ examples
    4. Expert learns from these interactions

    **Benefits:**

    * No manual data creation
    * Real-world examples
    * Continuous improvement
    * Natural workflow
  </Tab>

  <Tab title="Fine-Tuning">
    **Train custom models:**

    * Use marked QA pairs as training data
    * Create specialized model versions
    * Deploy fine-tuned models
    * A/B test different versions

    **Results:**

    * Consistent behavior
    * Domain expertise
    * Brand voice compliance
    * Better accuracy
  </Tab>

  <Tab title="Model Distillation">
    **Advanced technique:**

    * Use GPT-4 to generate perfect examples
    * Train GPT-3.5 on those examples
    * Get 90-95% of GPT-4's quality
    * At 10-50x lower cost

    **Use for:**

    * Production cost optimization
    * High-volume applications
    * Speed improvements
  </Tab>
</Tabs>

Learn more: [Fine-Tuning](/concepts/fine-tuning/introduction) | [Model Distillation](/concepts/fine-tuning/model-distillation)

### 3. Provider Keys Management

**Unified API key management:**

<AccordionGroup>
  <Accordion icon="robot" title="LLM Providers">
    Manage keys for language models:

    * OpenAI (GPT-4, GPT-3.5)
    * Anthropic (Claude 3)
    * Google (Gemini)
    * Mistral, DeepSeek, Cohere
    * Groq, Azure, Ollama
    * And 10+ more

    **Benefits:**

    * One place for all keys
    * Primary & fallback keys
    * Key validation
    * Usage tracking
  </Accordion>

  <Accordion icon="microphone" title="Voice Providers">
    Text-to-speech and speech-to-text:

    * OpenAI TTS & Whisper
    * ElevenLabs
    * Google TTS
    * Azure Speech
    * Browser native

    **Flexibility:**

    * Different keys for voice vs text
    * Multiple accounts
    * Separate billing
  </Accordion>

  <Accordion icon="image" title="Image Providers">
    Image generation services:

    * OpenAI DALL-E
    * Stability AI
    * Midjourney
    * BFL FLUX (profile pictures)

    **Uses:**

    * Expert profile pictures
    * Content generation
    * Visual assets
  </Accordion>
</AccordionGroup>

Learn more: [Provider Keys](/essentials/provider-keys)

### 4. Multimodal Capabilities

**Beyond text interactions:**

<CardGroup cols={2}>
  <Card title="Voice Input" icon="microphone">
    **Speech-to-Text:**

    * Real-time voice recording
    * Multiple languages
    * Hands-free operation
    * Mobile-friendly

    **Powered by:**

    * OpenAI Whisper
    * Google Speech-to-Text
    * Azure Speech
    * Browser native
  </Card>

  <Card title="Voice Output" icon="volume-high">
    **Text-to-Speech:**

    * Natural-sounding voices
    * Multiple voice options
    * Streaming audio
    * Auto-play settings

    **Powered by:**

    * OpenAI TTS (6 voices)
    * ElevenLabs (100+ voices)
    * Google TTS (WaveNet)
    * Azure Neural Voices
  </Card>

  <Card title="Image Processing" icon="image">
    **Visual AI:**

    * AI-generated profile pics
    * Image analysis
    * Screenshot understanding
    * Diagram explanations

    **Capabilities:**

    * OCR (text extraction)
    * Object recognition
    * Visual Q\&A
    * Image generation
  </Card>

  <Card title="File Handling" icon="file">
    **Document Processing:**

    * PDF, Word, Excel
    * Code files
    * CSV data analysis
    * JSON/XML parsing

    **Operations:**

    * Read and analyze
    * Extract information
    * Generate summaries
    * Answer questions
  </Card>
</CardGroup>

Learn more: [Modalities](/essentials/modalities) | [Multimodal AI](/concepts/multimodal/introduction)

### 5. DeepAgents Workspace

**Advanced autonomous capabilities:**

<Tabs>
  <Tab title="Task Management">
    **Hierarchical todos:**

    ```
    Main Goal
    ├─ Subtask 1
    │  ├─ Step 1.1
    │  └─ Step 1.2
    └─ Subtask 2
       └─ Step 2.1
    ```

    **Features:**

    * Automatic task breakdown
    * Status tracking (pending, in\_progress, completed)
    * Progress monitoring
    * Priority management

    **Use for:**

    * Multi-step projects
    * Complex workflows
    * Autonomous execution
  </Tab>

  <Tab title="File System">
    **Virtual workspace:**

    Each conversation has persistent files:

    * Code files
    * Documents
    * Data files
    * Reports
    * Outputs

    **Operations:**

    * Create/edit files
    * Organize by type
    * Download anytime
    * Persist across sessions

    **Use for:**

    * Development projects
    * Data analysis
    * Report generation
    * Content creation
  </Tab>

  <Tab title="Human-in-the-Loop">
    **Control automation:**

    **Auto Mode:**

    ```
    Agent plans → Executes → Completes
    ```

    **HITL Mode:**

    ```
    Agent plans → Asks approval → If yes, executes
    ```

    **When to use HITL:**

    * Critical operations
    * Database modifications
    * Financial transactions
    * Learning agent behavior
  </Tab>
</Tabs>

Learn more: [DeepAgents Workspace](/essentials/deepagents-workspace) | [DeepAgents Concepts](/concepts/deepagents/introduction)

### 6. Distribution Channels

**Deploy your experts everywhere:**

<Tabs>
  <Tab title="B-Bot Chat">
    **Internal chat application:**

    * Direct integration
    * Full feature access
    * User management
    * Team collaboration
    * Real-time conversations

    **Best for:**

    * Internal teams
    * Secure environments
    * Full-featured usage
    * Power users
  </Tab>

  <Tab title="Embed Widget">
    **Website integration:**

    * Customizable chat widget
    * Brand matching (colors, icons)
    * Multiple styles (bubble, sidebar, fullscreen)
    * Easy integration (copy/paste script)
    * Mobile responsive

    **Best for:**

    * Customer support
    * Website visitors
    * Lead generation
    * Public access
  </Tab>

  <Tab title="API">
    **Custom integrations:**

    * REST endpoints
    * Streaming responses
    * Full control
    * Authentication
    * Webhook support

    **Best for:**

    * Mobile apps
    * Custom applications
    * System integrations
    * Developers
  </Tab>

  <Tab title="Tasks">
    **Scheduled automation:**

    * Cron scheduling
    * Automated reports
    * Data processing
    * Monitoring
    * Background jobs

    **Best for:**

    * Recurring reports
    * Batch processing
    * Monitoring
    * Automation
  </Tab>

  <Tab title="Templates">
    **Public sharing:**

    * Template marketplace
    * Share your experts
    * Help community
    * Build reputation
    * Get featured

    **Best for:**

    * Community contribution
    * Showcasing capabilities
    * Building brand
    * Helping others
  </Tab>
</Tabs>

Learn more: [Distribution Channels](/essentials/distribution-channels)

### 7. App Integrations

**Connect to external services:**

<CardGroup cols={3}>
  <Card title="Google Workspace" icon="google">
    * Gmail
    * Drive
    * Calendar
    * Sheets
    * Docs
  </Card>

  <Card title="Communication" icon="comments">
    * Slack
    * Discord
    * Microsoft Teams
    * WhatsApp (coming soon)
  </Card>

  <Card title="Development" icon="code">
    * GitHub
    * GitLab
    * Bitbucket
    * Jira
    * Linear
  </Card>

  <Card title="Productivity" icon="list-check">
    * Notion
    * Trello
    * Asana
    * Monday.com
    * Airtable
  </Card>

  <Card title="CRM & Sales" icon="handshake">
    * Salesforce
    * HubSpot
    * Pipedrive
    * Zendesk
  </Card>

  <Card title="Data & Analytics" icon="chart-line">
    * Databases
    * APIs
    * Webhooks
    * Analytics tools
  </Card>
</CardGroup>

**Each expert can:**

* Use different apps
* Have unique credentials
* Connect multiple services
* Automate workflows

Learn more: [Apps & Integrations](/essentials/apps)

## Common Workflows

### Workflow 1: Customer Support Expert

```
1. Create Expert
   ├─ Name: "Support Assistant"
   ├─ Profession: "Customer Support Specialist"
   └─ Abilities: Handle inquiries, troubleshoot, escalate

2. Fine-Tune
   ├─ Mark 200 perfect support conversations
   └─ Train custom model

3. Connect Apps
   ├─ Zendesk (ticket system)
   ├─ Knowledge base
   └─ CRM

4. Deploy via Embed
   └─ Add chat widget to website

5. Monitor & Improve
   ├─ Track satisfaction
   ├─ Mark new good examples
   └─ Retrain monthly
```

### Workflow 2: Data Analysis Expert

```
1. Create Expert
   ├─ Name: "Data Analyst AI"
   ├─ Enable DeepAgents mode
   └─ Configure file handling

2. Connect Apps
   ├─ Database connections
   ├─ Google Sheets
   └─ Visualization tools

3. Deploy via Tasks
   ├─ Schedule: Daily at 8 AM
   ├─ Input: Latest data
   └─ Output: Analysis report

4. Use Workspace
   ├─ Generate analysis files
   ├─ Create visualizations
   └─ Write summary reports
```

### Workflow 3: Content Creation Expert

```
1. Create Expert
   ├─ Fine-tune on brand voice
   └─ Connect to content tools

2. Use Model Distillation
   ├─ Generate examples with GPT-4
   ├─ Train GPT-3.5
   └─ 93% cost reduction

3. Deploy via API
   └─ Integrate with CMS

4. Multi-Channel
   ├─ Social media posts
   ├─ Blog drafts
   └─ Email campaigns
```

## Getting Started Checklist

<Steps>
  <Step title="Set Up Provider Keys">
    Add API keys for the services you'll use

    * OpenAI, Anthropic, or others
    * Voice providers (optional)
    * Image providers (optional)

    → [Provider Keys Setup](/essentials/provider-keys)
  </Step>

  <Step title="Create Your First Expert">
    Follow the 8-step creation wizard

    * Define personality
    * Choose model
    * Configure capabilities

    → [Create Expert Guide](/essentials/create-expert)
  </Step>

  <Step title="Test in Chat">
    Have conversations to refine behavior

    * Test different scenarios
    * Mark good responses (QA marking)
    * Adjust abilities as needed

    → [Chat Features](/essentials/chat)
  </Step>

  <Step title="Optional: Fine-Tune">
    Train on your specific use case

    * Collect 50-500 examples
    * Train custom model
    * Deploy fine-tuned version

    → [Training Guide](/essentials/training)
  </Step>

  <Step title="Deploy via Distribution Channel">
    Make your expert available to users

    * Choose channel (Chat, Embed, API, Tasks)
    * Configure settings
    * Launch!

    → [Distribution Channels](/essentials/distribution-channels)
  </Step>

  <Step title="Monitor & Improve">
    Continuous improvement cycle

    * Collect user feedback
    * Mark new examples
    * Retrain periodically
    * Update configuration
  </Step>
</Steps>

## Next Steps

<CardGroup cols={2}>
  <Card title="Create Your Expert" icon="plus" href="/essentials/create-expert">
    Start building your first AI expert
  </Card>

  <Card title="Fine-Tuning" icon="brain" href="/concepts/fine-tuning/introduction">
    Learn about training custom models
  </Card>

  <Card title="DeepAgents" icon="robot" href="/concepts/deepagents/introduction">
    Understand autonomous capabilities
  </Card>

  <Card title="Distribution" icon="share-nodes" href="/essentials/distribution-channels">
    Deploy your expert everywhere
  </Card>
</CardGroup>