Real-time Streaming - Burki Voice AI Docs

All Burki SDKs support real-time WebSocket streaming for live transcripts during calls and campaign progress monitoring. This enables building powerful real-time dashboards, monitoring tools, and analytics systems.

Overview

Burki provides two real-time streaming endpoints:

Feature	Description	Use Cases
Live Transcripts	Stream real-time transcriptions during active calls	Call monitoring, live captions, sentiment analysis
Campaign Progress	Stream campaign execution updates	Progress dashboards, real-time analytics, alerting

Live Transcript Streaming

Stream real-time transcripts as they are generated during calls. Perfect for building:

Call monitoring dashboards
Live supervisor interfaces
Real-time sentiment analysis
Accessibility features (live captions)
Training and coaching tools

Event Types

Event Type	Description
`connection_established`	WebSocket connection successful
`transcript`	New transcript segment from user or assistant
`call_status`	Call status change (in-progress, completed, failed)
`error`	Error message
`pong`	Response to ping (keepalive)

Transcript Event Data

{
  "type": "transcript",
  "call_sid": "CA123abc...",
  "timestamp": "2026-01-19T12:00:05.123Z",
  "data": {
    "content": "Hello, how can I help you today?",
    "speaker": "assistant",
    "is_final": true,
    "confidence": 0.95,
    "segment_start": 5.2,
    "segment_end": 7.1,
    "created_at": "2026-01-19T12:00:05.123Z"
  }
}

Implementation Examples

Python
JavaScript
Go

The Python SDK uses async context managers for streaming:

import asyncio
from burki import BurkiClient

async def monitor_call(call_sid: str):
    client = BurkiClient(api_key="your-api-key")
    
    async with client.realtime.live_transcript(call_sid=call_sid) as stream:
        async for event in stream:
            if event.type == "transcript":
                speaker = event.data.speaker
                content = event.data.content
                print(f"[{speaker}]: {content}")
                
            elif event.type == "call_status":
                print(f"Call status: {event.status}")
                if event.status == "completed":
                    break
                    
            elif event.type == "error":
                print(f"Error: {event.message}")
                break

# Run the monitor
asyncio.run(monitor_call("CA123abc..."))

Features:

Automatic connection management with context manager
Async iteration for events
Clean disconnection on exit

The JavaScript SDK uses async iterators:

import { BurkiClient } from '@burki.dev/sdk';

async function monitorCall(callSid: string) {
  const client = new BurkiClient({ apiKey: 'your-api-key' });
  
  const stream = client.realtime.liveTranscript(callSid);
  await stream.connect();

  try {
    for await (const event of stream) {
      switch (event.type) {
        case 'transcript':
          console.log(`[${event.speaker}]: ${event.content}`);
          break;
          
        case 'call_status':
          console.log(`Call status: ${event.status}`);
          if (event.status === 'completed') {
            stream.disconnect();
            return;
          }
          break;
          
        case 'error':
          console.error(`Error: ${event.message}`);
          stream.disconnect();
          return;
      }
    }
  } finally {
    stream.disconnect();
  }
}

monitorCall('CA123abc...');

Features:

Async iterator pattern (for await…of)
Explicit connect/disconnect control
TypeScript type safety

The Go SDK uses channels for event streaming:

package main

import (
    "fmt"
    "log"
    "time"

    burki "github.com/burki-ai/burki-go/burki"
)

func monitorCall(callSID string) {
    client := burki.NewClient("your-api-key")
    
    stream := client.Realtime.LiveTranscript(callSID)
    if err := stream.Connect(); err != nil {
        log.Fatal(err)
    }
    defer stream.Close()

    // Keepalive goroutine
    go func() {
        ticker := time.NewTicker(30 * time.Second)
        defer ticker.Stop()
        for range ticker.C {
            stream.SendPing()
        }
    }()

    // Event loop using select
    for {
        select {
        case event := <-stream.Events:
            switch e := event.(type) {
            case *burki.TranscriptEvent:
                fmt.Printf("[%s]: %s\n", e.Speaker, e.Content)
            case *burki.CallStatusEvent:
                fmt.Printf("Call status: %s\n", e.Status)
                if e.Status == "completed" {
                    return
                }
            }
        case err := <-stream.Errors:
            log.Printf("Error: %v\n", err)
            return
        case <-stream.Done:
            fmt.Println("Stream closed")
            return
        }
    }
}

Features:

Channel-based event delivery
Select statement for concurrent handling
Explicit goroutine management

Campaign Progress Streaming

Monitor campaign execution in real-time. Perfect for building:

Campaign progress dashboards
Real-time analytics displays
Automated alerting systems
Performance monitoring tools

Event Types

Event Type	Description
`connection_established`	WebSocket connection successful
`progress`	Periodic progress update with counts
`contact_completed`	Individual contact call completed
`campaign_completed`	Campaign finished (all contacts processed)
`error`	Error message

Progress Event Data

{
  "type": "progress",
  "campaign_id": 123,
  "timestamp": "2026-01-19T12:05:00.000Z",
  "data": {
    "total_contacts": 100,
    "completed_contacts": 45,
    "successful_contacts": 38,
    "failed_contacts": 7,
    "pending_contacts": 55,
    "success_rate": 84.4
  }
}

Contact Completed Event

{
  "type": "contact_completed",
  "campaign_id": 123,
  "timestamp": "2026-01-19T12:05:30.000Z",
  "data": {
    "phone_number": "+14155551234",
    "name": "John Smith",
    "outcome": "success",
    "duration": 125,
    "call_sid": "CA456def..."
  }
}

Implementation Examples

Python
JavaScript
Go

import asyncio
from burki import BurkiClient

async def monitor_campaign(campaign_id: int):
    client = BurkiClient(api_key="your-api-key")
    
    async with client.realtime.campaign_progress(campaign_id=campaign_id) as stream:
        async for event in stream:
            if event.type == "progress":
                completed = event.data.completed_contacts
                total = event.data.total_contacts
                rate = event.data.success_rate
                print(f"Progress: {completed}/{total} ({rate:.1f}% success)")
                
            elif event.type == "contact_completed":
                phone = event.data.phone_number
                outcome = event.data.outcome
                print(f"Completed: {phone} - {outcome}")
                
            elif event.type == "campaign_completed":
                print(f"\nCampaign finished!")
                print(f"Total contacts: {event.data.total_contacts}")
                print(f"Success rate: {event.data.success_rate:.1f}%")
                break

# Run the monitor
asyncio.run(monitor_campaign(123))

import { BurkiClient } from '@burki.dev/sdk';

async function monitorCampaign(campaignId: number) {
  const client = new BurkiClient({ apiKey: 'your-api-key' });
  
  const stream = client.realtime.campaignProgress(campaignId);
  await stream.connect();

  for await (const event of stream) {
    switch (event.type) {
      case 'progress':
        console.log(
          `Progress: ${event.completedContacts}/${event.totalContacts} ` +
          `(${event.successRate.toFixed(1)}% success)`
        );
        break;
        
      case 'contact_completed':
        console.log(`Completed: ${event.phoneNumber} - ${event.outcome}`);
        break;
        
      case 'campaign_completed':
        console.log('\nCampaign finished!');
        console.log(`Total contacts: ${event.totalContacts}`);
        console.log(`Success rate: ${event.successRate.toFixed(1)}%`);
        stream.disconnect();
        return;
    }
  }
}

monitorCampaign(123);

package main

import (
    "fmt"
    "log"
    "time"

    burki "github.com/burki-ai/burki-go/burki"
)

func monitorCampaign(campaignID int) {
    client := burki.NewClient("your-api-key")
    
    stream := client.Realtime.CampaignProgress(campaignID)
    if err := stream.Connect(); err != nil {
        log.Fatal(err)
    }
    defer stream.Close()

    // Keepalive
    go func() {
        ticker := time.NewTicker(30 * time.Second)
        defer ticker.Stop()
        for range ticker.C {
            stream.SendPing()
        }
    }()

    for {
        select {
        case event := <-stream.Events:
            switch e := event.(type) {
            case *burki.CampaignProgressEvent:
                fmt.Printf("Progress: %d/%d (%.1f%% success)\n",
                    e.CompletedContacts, e.TotalContacts, e.SuccessRate)
            case *burki.ContactCompletedEvent:
                fmt.Printf("Completed: %s - %s\n", e.PhoneNumber, e.Outcome)
            case *burki.CampaignCompletedEvent:
                fmt.Printf("\nCampaign finished!\n")
                fmt.Printf("Total contacts: %d\n", e.TotalContacts)
                fmt.Printf("Success rate: %.1f%%\n", e.SuccessRate)
                return
            }
        case err := <-stream.Errors:
            log.Printf("Error: %v\n", err)
            return
        case <-stream.Done:
            return
        }
    }
}

Connection Management

Keepalive / Ping

Send periodic pings to keep the WebSocket connection alive, especially for long-running sessions:

import asyncio

async def with_keepalive(stream, interval=30):
    async def ping_loop():
        while True:
            await asyncio.sleep(interval)
            await stream.send_ping()
    
    # Start ping task
    ping_task = asyncio.create_task(ping_loop())
    
    try:
        async for event in stream:
            yield event
    finally:
        ping_task.cancel()

Reconnection Strategy

Implement exponential backoff for automatic reconnection:

import asyncio
from burki import BurkiClient

async def resilient_stream(call_sid: str, max_retries=5):
    client = BurkiClient(api_key="your-api-key")
    retries = 0
    
    while retries < max_retries:
        try:
            async with client.realtime.live_transcript(call_sid) as stream:
                retries = 0  # Reset on successful connection
                async for event in stream:
                    yield event
                    
        except Exception as e:
            retries += 1
            if retries >= max_retries:
                raise
            
            # Exponential backoff
            delay = min(2 ** retries, 60)
            print(f"Reconnecting in {delay}s (attempt {retries})")
            await asyncio.sleep(delay)

Best Practices

When to Use Streaming vs Polling

Use Streaming When:

You need real-time updates (< 1 second latency)
Building live dashboards or monitoring interfaces
Processing events as they happen (e.g., sentiment analysis)
Monitoring long-running operations (campaigns, calls)

Use Polling When:

Periodic status checks are sufficient (> 30 seconds)
Simple status display without live updates
Lower resource usage is a priority
Connection stability is a concern

Resource Management

Always Clean Up:

Close streams when done (close(), disconnect(), or context manager exit)
Cancel keepalive tasks/goroutines
Handle disconnection events gracefully

Memory Considerations:

Don’t store all events in memory for long streams
Process or persist events as they arrive
Use bounded buffers if accumulating data

Connection Limits:

Maximum 10 concurrent connections per call
Maximum 100 messages per second per connection
Consider connection pooling for multiple streams

Error Handling

Handle All Error Types:

Connection errors (network issues)
Authentication errors (invalid API key)
Not found errors (invalid call_sid or campaign_id)
Rate limit errors (too many connections)

Graceful Degradation:

Fall back to polling if streaming fails
Show stale data with “updating…” indicator
Log errors for debugging

User Experience:

Show connection status to users
Indicate when data may be stale
Provide manual refresh option

Security Considerations

API Key Protection:

Never expose API keys in client-side code (browser)
Use server-side proxies for browser applications
Store keys in environment variables

Data Privacy:

Transcripts may contain sensitive information
Implement appropriate access controls
Consider data retention policies

Connection Security:

Always use WSS (WebSocket Secure) in production
Validate SSL certificates
Implement authentication tokens if needed

Use Case Examples

Call Monitoring Dashboard

Build a real-time call center monitoring interface:

class CallMonitoringDashboard {
  private activeStreams = new Map<string, RealtimeStream>();
  
  async monitorCall(callSid: string, displayElement: HTMLElement) {
    const stream = client.realtime.liveTranscript(callSid);
    await stream.connect();
    
    this.activeStreams.set(callSid, stream);
    
    for await (const event of stream) {
      if (event.type === 'transcript') {
        this.updateTranscriptDisplay(displayElement, event);
      } else if (event.type === 'call_status' && event.status === 'completed') {
        this.stopMonitoring(callSid);
        break;
      }
    }
  }
  
  stopMonitoring(callSid: string) {
    const stream = this.activeStreams.get(callSid);
    if (stream) {
      stream.disconnect();
      this.activeStreams.delete(callSid);
    }
  }
  
  stopAll() {
    for (const [callSid, stream] of this.activeStreams) {
      stream.disconnect();
    }
    this.activeStreams.clear();
  }
}

Real-time Sentiment Analysis

Analyze call sentiment as transcripts arrive:

from textblob import TextBlob

async def analyze_sentiment(call_sid: str):
    client = BurkiClient(api_key="your-api-key")
    sentiment_scores = []
    
    async with client.realtime.live_transcript(call_sid) as stream:
        async for event in stream:
            if event.type == "transcript" and event.data.speaker == "user":
                # Analyze user sentiment
                blob = TextBlob(event.data.content)
                score = blob.sentiment.polarity
                sentiment_scores.append(score)
                
                # Alert on negative sentiment
                if score < -0.5:
                    await alert_supervisor(call_sid, event.data.content, score)
                
                # Calculate rolling average
                avg = sum(sentiment_scores[-10:]) / min(len(sentiment_scores), 10)
                print(f"Rolling sentiment: {avg:.2f}")

Campaign Progress Webhook

Forward campaign events to external systems:

func forwardCampaignEvents(campaignID int, webhookURL string) {
    client := burki.NewClient(os.Getenv("BURKI_API_KEY"))
    stream := client.Realtime.CampaignProgress(campaignID)
    
    if err := stream.Connect(); err != nil {
        log.Fatal(err)
    }
    defer stream.Close()

    httpClient := &http.Client{Timeout: 10 * time.Second}
    
    for {
        select {
        case event := <-stream.Events:
            // Forward to webhook
            jsonData, _ := json.Marshal(event)
            resp, err := httpClient.Post(webhookURL, "application/json", 
                bytes.NewBuffer(jsonData))
            if err != nil {
                log.Printf("Webhook error: %v", err)
            } else {
                resp.Body.Close()
            }
            
        case <-stream.Done:
            return
        }
    }
}

Next Steps

Python SDK

Full Python SDK documentation

JavaScript SDK

Full JavaScript/TypeScript SDK documentation

Go SDK

Full Go SDK documentation

Live Transcript API

REST API WebSocket reference

SDK Documentation

Documentation Index

​Overview

​Live Transcript Streaming

​Event Types

​Transcript Event Data

​Implementation Examples

​Campaign Progress Streaming

​Event Types

​Progress Event Data

​Contact Completed Event

​Implementation Examples

​Connection Management

​Keepalive / Ping

​Reconnection Strategy

​Best Practices

​Use Case Examples

​Call Monitoring Dashboard

​Real-time Sentiment Analysis

​Campaign Progress Webhook

​Next Steps

Python SDK

JavaScript SDK

Go SDK

Live Transcript API

Overview

Live Transcript Streaming

Event Types

Transcript Event Data

Implementation Examples

Campaign Progress Streaming

Event Types

Progress Event Data

Contact Completed Event

Implementation Examples

Connection Management

Keepalive / Ping

Reconnection Strategy

Best Practices

Use Case Examples

Call Monitoring Dashboard

Real-time Sentiment Analysis

Campaign Progress Webhook

Next Steps