> ## Documentation Index
> Fetch the complete documentation index at: https://docs.burki.dev/llms.txt
> Use this file to discover all available pages before exploring further.
# Core Concepts & Architecture
> Understanding how Burki Voice AI works
Understanding these core concepts will help you get the most out of Burki Voice AI!
***
## What is an Assistant?
An **Assistant** is a virtual agent that can answer calls, talk to users, and perform tasks using AI. Each assistant has its own configuration for language, voice, hearing, and integrations.
* You can create multiple assistants for different use cases (e.g., sales, support, info lines).
* Each assistant can have its own settings, phone number, and knowledge base.
* Assistants can be updated in real-time without any downtime.
***
## Multi-Tenancy & Organizations
* **Organizations:** Group your assistants, team members, and resources under one account. Great for agencies, teams, or businesses with many brands.
* **Role-Based Access:** Admins, members, and viewers each have different permissions.
* **Resource Isolation:** Each organization's data is completely separate and secure.
***
## Provider Modes
Burki offers flexibility in how you connect to AI providers:
We manage all provider connections. You select providers and models, we handle the rest.
Connect your own API keys. You pay providers directly and maintain full control.
* **Cost Control:** Use your existing enterprise pricing with BYO mode
* **Try New Models:** Test new AI models as they launch
* **Optimize:** Use Burki Cloud for convenience, BYO for specific needs
* **Fallbacks:** Set up backup providers for reliability
***
## How a Call Flows Through the System
When someone calls your Burki phone number, here's what happens:
```mermaid theme={null}
graph TD
A[Caller dials number] --> B[Telephony Provider]
B --> C[Burki Platform]
C --> D[STT: Speech to Text]
D --> E[LLM: Generate Response]
E --> F[TTS: Text to Speech]
F --> G[Caller hears response]
E -->|"If tool needed"| H[Execute Tool]
H --> E
E -->|"If primary fails"| I[Fallback LLM]
I --> F
```
### Component Breakdown
| Component | What It Does | Providers |
| ------------- | --------------------------------------------- | ---------------------------------------------------------------------------------------------- |
| **Telephony** | Handles phone calls and audio streaming | Twilio, Vonage, Telnyx |
| **STT** | Converts caller's speech to text | Deepgram, ElevenLabs, Azure, Uplift, Speechmatics, Telnyx, Soniox |
| **LLM** | Understands context and generates responses | OpenAI, Azure OpenAI, Anthropic, Gemini, Groq, xAI, custom HTTP/WebSocket |
| **TTS** | Converts AI response back to speech | ElevenLabs, Deepgram, Cartesia, Azure, Inworld, Resemble, OpenAI, Kokoro, Uplift, Murf, Soniox |
| **Tools** | Executes actions (transfers, API calls, etc.) | Built-in + Custom |
You can customize every step—choose your favorite providers, set up fallbacks, and fine-tune the experience!
***
## Configuration Sections
When setting up an assistant, you'll configure these sections:
| Section | What You Configure |
| ------------------------ | ---------------------------------------- |
| **Basic Info** | Name, description, and greeting message |
| **AI Configuration** | LLM, TTS, and STT providers and models |
| **RAG (Knowledge Base)** | Documents and web pages for context |
| **Call Management** | Interruptions, recordings, and call flow |
| **Tools** | Actions your assistant can take |
| **Phone Numbers** | Which numbers route to this assistant |
***
## System Architecture
```mermaid theme={null}
graph TD
subgraph users [Users]
Web[Web Dashboard]
API[REST API]
Phone[Phone Calls]
end
subgraph platform [Burki Platform]
Gateway[API Gateway]
Core[Core Engine]
Audio[Audio Processing]
Tools[Tool Execution]
end
subgraph providers [AI Providers]
STT[Speech-to-Text]
LLM[Language Models]
TTS[Text-to-Speech]
end
subgraph telephony [Telephony]
Twilio[Twilio]
Vonage[Vonage]
Telnyx[Telnyx]
end
subgraph storage [Data]
DB[(Database)]
Files[(File Storage)]
end
Web --> Gateway
API --> Gateway
Phone --> telephony
telephony --> Gateway
Gateway --> Core
Core --> Audio
Core --> Tools
Audio --> STT
Audio --> TTS
Core --> LLM
Core --> DB
Core --> Files
```
### Key Components
* **API Gateway**: Handles authentication, rate limiting, and request routing
* **Core Engine**: Manages conversation state, assistant configuration, and business logic
* **Audio Processing**: Real-time audio streaming with noise reduction
* **Tool Execution**: Secure execution environment for custom actions
* **Storage**: Encrypted storage for conversations, recordings, and configuration
***
## Data Flow
### Inbound Call
1. **Call arrives** at your phone number via telephony provider
2. **Audio streams** to Burki in real-time over WebSocket
3. **STT converts** speech to text with sub-second latency
4. **LLM generates** a contextual response
5. **TTS converts** response to audio
6. **Audio streams** back to the caller
7. **Call data** is logged for analytics and transcripts
### Outbound Call (API-Initiated)
1. **API request** triggers an outbound call
2. **Burki connects** via telephony provider
3. **Same flow** as inbound once connected
4. **Webhook** notifies your system of call status
***
## Security & Compliance
* **Encryption**: All data encrypted at rest (AES-256) and in transit (TLS 1.3)
* **API Key Security**: Keys are hashed and never stored in plain text
* **Provider Isolation**: BYO keys are encrypted and only used for your requests
* **Audit Logs**: Complete audit trail of all system access
* **Data Retention**: Configurable retention policies for recordings and transcripts
***
## Next Steps
Learn about LLM, TTS, and STT configuration options
Connect with external services and APIs