Tutorial 2: Streaming Responses

Learn how to implement real-time streaming for better user experience.

What You’ll Learn

• Stream LLM responses in real-time
• Process chunks as they arrive
• Handle streaming errors
• Track progress and completion

Why Streaming?

Streaming provides immediate feedback:

Without Streaming:

[3 second wait...]
Here's a complete response about streaming...

With Streaming:

Here's→ a→ complete→ response→ about→ streaming...

Users see results immediately and can stop generation if needed.

Prerequisites

Complete Tutorial 1: Your First Conversation or understand basic SDK usage.

Basic Streaming

Use conv.Stream() instead of conv.Send():

package main

import (
    "context"
    "fmt"
    "log"

    "github.com/AltairaLabs/PromptKit/sdk"
)

func main() {
    conv, err := sdk.Open("./hello.pack.json", "chat")
    if err != nil {
        log.Fatal(err)
    }
    defer conv.Close()

    ctx := context.Background()

    fmt.Print("Assistant: ")
    for chunk := range conv.Stream(ctx, "Tell me a short story") {
        if chunk.Error != nil {
            log.Printf("Error: %v", chunk.Error)
            break
        }
        if chunk.Type == sdk.ChunkDone {
            fmt.Println("\n[Done]")
            break
        }
        fmt.Print(chunk.Text)
    }
}

Understanding Stream Chunks

Each chunk contains:

type StreamChunk struct {
    Type  ChunkType // ChunkText, ChunkToolCall, ChunkDone
    Text  string    // Text content (for ChunkText)
    Error error     // Non-nil if error occurred
}

Chunk Types

• sdk.ChunkText - Text content arrived
• sdk.ChunkToolCall - Tool is being called
• sdk.ChunkDone - Stream completed

Collecting Full Response

Track the complete response while streaming:

var fullText string

for chunk := range conv.Stream(ctx, "Write a poem") {
    if chunk.Error != nil {
        log.Printf("Error: %v", chunk.Error)
        break
    }
    if chunk.Type == sdk.ChunkDone {
        break
    }

    fmt.Print(chunk.Text)  // Real-time display
    fullText += chunk.Text  // Collect for later
}

fmt.Printf("\n\nTotal length: %d characters\n", len(fullText))

Progress Tracking

Show progress indicators:

charCount := 0

for chunk := range conv.Stream(ctx, "Tell me about AI") {
    if chunk.Error != nil {
        break
    }
    if chunk.Type == sdk.ChunkDone {
        fmt.Printf("\n\n[Complete - %d characters]\n", charCount)
        break
    }

    fmt.Print(chunk.Text)
    charCount += len(chunk.Text)
}

Interactive Streaming Chat

Build a streaming chatbot:

package main

import (
    "bufio"
    "context"
    "fmt"
    "log"
    "os"
    "strings"

    "github.com/AltairaLabs/PromptKit/sdk"
)

func main() {
    conv, err := sdk.Open("./hello.pack.json", "chat")
    if err != nil {
        log.Fatal(err)
    }
    defer conv.Close()

    ctx := context.Background()
    scanner := bufio.NewScanner(os.Stdin)

    fmt.Println("Streaming chat ready! Type 'quit' to exit.")

    for {
        fmt.Print("\nYou: ")
        if !scanner.Scan() {
            break
        }

        msg := strings.TrimSpace(scanner.Text())
        if msg == "quit" {
            break
        }

        fmt.Print("Assistant: ")
        for chunk := range conv.Stream(ctx, msg) {
            if chunk.Error != nil {
                log.Printf("\nError: %v", chunk.Error)
                break
            }
            if chunk.Type == sdk.ChunkDone {
                fmt.Println()
                break
            }
            fmt.Print(chunk.Text)
        }
    }
}

Error Handling

Handle errors gracefully:

for chunk := range conv.Stream(ctx, "Generate content") {
    if chunk.Error != nil {
        // Check error type
        if errors.Is(chunk.Error, context.DeadlineExceeded) {
            fmt.Println("\n[Timeout - response truncated]")
        } else {
            fmt.Printf("\n[Error: %v]", chunk.Error)
        }
        break
    }

    if chunk.Type == sdk.ChunkDone {
        break
    }

    fmt.Print(chunk.Text)
}

Timeout with Context

Set a streaming timeout:

ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
defer cancel()

for chunk := range conv.Stream(ctx, "Tell me a long story") {
    if chunk.Error != nil {
        break
    }
    if chunk.Type == sdk.ChunkDone {
        break
    }
    fmt.Print(chunk.Text)
}

What You’ve Learned

✅ Stream responses with conv.Stream()
✅ Process chunks as they arrive
✅ Track progress and completion
✅ Handle streaming errors
✅ Build interactive streaming apps

Next Steps

• Tutorial 3: Tools - Add function calling
• Tutorial 4: Variables - Template variables

Complete Example

See the full example at sdk/examples/streaming/.