ANDROID_STREAMING_GUIDE.md

ChatBIA 안드로이드 스트리밍 연동 가이드

📡 API 엔드포인트

1. 일반 채팅 (비스트리밍)

POST /chat

• 타임아웃 위험: 긴 응답 시 타임아웃 발생 가능
• 안드로이드에서 권장하지 않음

2. 스트리밍 채팅 ✅ 권장

POST /chat/stream

• 타임아웃 방지: 토큰 단위로 실시간 수신
• 안드로이드에 최적화
• SSE (Server-Sent Events) 방식

---

🔧 안드로이드 구현 (Kotlin)

1. build.gradle 의존성 추가

dependencies {
    // OkHttp for SSE streaming
    implementation("com.squareup.okhttp3:okhttp:4.12.0")
    implementation("com.squareup.okhttp3:okhttp-sse:4.12.0")

    // JSON 파싱
    implementation("com.google.code.gson:gson:2.10.1")

    // Coroutines
    implementation("org.jetbrains.kotlinx:kotlinx-coroutines-android:1.7.3")
}

2. 데이터 모델

// ChatRequest.kt
data class ChatRequest(
    val message: String,
    val mode: String = "bsl",  // "bsl" or "general"
    val max_tokens: Int = 1024,
    val temperature: Float = 0.7f
)

// StreamingResponse.kt
data class StreamingResponse(
    val token: String = "",
    val done: Boolean = false,
    val token_count: Int = 0,
    val mode: String = "",
    val error: String? = null
)

3. ChatBIA API 클라이언트

// ChatBiaApiClient.kt
import com.google.gson.Gson
import kotlinx.coroutines.flow.Flow
import kotlinx.coroutines.flow.flow
import okhttp3.*
import okhttp3.MediaType.Companion.toMediaType
import okhttp3.RequestBody.Companion.toRequestBody
import okhttp3.sse.EventSource
import okhttp3.sse.EventSourceListener
import okhttp3.sse.EventSources
import kotlin.coroutines.resume
import kotlin.coroutines.resumeWithException
import kotlin.coroutines.suspendCoroutine

class ChatBiaApiClient(private val baseUrl: String) {

    private val client = OkHttpClient.Builder()
        .connectTimeout(30, TimeUnit.SECONDS)
        .readTimeout(60, TimeUnit.SECONDS)  // 스트리밍은 긴 타임아웃
        .writeTimeout(30, TimeUnit.SECONDS)
        .build()

    private val gson = Gson()

    /**
     * 스트리밍 채팅 (권장)
     * Flow를 통해 토큰 단위로 실시간 수신
     */
    fun chatStream(request: ChatRequest): Flow<StreamingResponse> = flow {
        suspendCoroutine<Unit> { continuation ->
            val url = "$baseUrl/chat/stream"

            // JSON 요청 body
            val jsonBody = gson.toJson(request)
            val requestBody = jsonBody.toRequestBody("application/json".toMediaType())

            val httpRequest = Request.Builder()
                .url(url)
                .post(requestBody)
                .addHeader("Accept", "text/event-stream")
                .build()

            // SSE EventSource 생성
            val eventSource = EventSources.createFactory(client)
                .newEventSource(httpRequest, object : EventSourceListener() {

                    override fun onOpen(eventSource: EventSource, response: Response) {
                        // 연결 성공
                    }

                    override fun onEvent(
                        eventSource: EventSource,
                        id: String?,
                        type: String?,
                        data: String
                    ) {
                        try {
                            val response = gson.fromJson(data, StreamingResponse::class.java)

                            // Flow로 emit
                            trySend(response)

                            // 완료 시 연결 종료
                            if (response.done) {
                                eventSource.cancel()
                                continuation.resume(Unit)
                            }

                        } catch (e: Exception) {
                            eventSource.cancel()
                            continuation.resumeWithException(e)
                        }
                    }

                    override fun onFailure(
                        eventSource: EventSource,
                        t: Throwable?,
                        response: Response?
                    ) {
                        continuation.resumeWithException(
                            t ?: Exception("SSE 연결 실패: ${response?.code}")
                        )
                    }

                    override fun onClosed(eventSource: EventSource) {
                        if (!continuation.isCompleted) {
                            continuation.resume(Unit)
                        }
                    }
                })
        }
    }

    /**
     * 일반 채팅 (비스트리밍)
     * 긴 응답 시 타임아웃 위험 있음
     */
    suspend fun chat(request: ChatRequest): ChatResponse = suspendCoroutine { continuation ->
        val url = "$baseUrl/chat"

        val jsonBody = gson.toJson(request)
        val requestBody = jsonBody.toRequestBody("application/json".toMediaType())

        val httpRequest = Request.Builder()
            .url(url)
            .post(requestBody)
            .build()

        client.newCall(httpRequest).enqueue(object : Callback {
            override fun onFailure(call: Call, e: IOException) {
                continuation.resumeWithException(e)
            }

            override fun onResponse(call: Call, response: Response) {
                if (response.isSuccessful) {
                    val body = response.body?.string()
                    val chatResponse = gson.fromJson(body, ChatResponse::class.java)
                    continuation.resume(chatResponse)
                } else {
                    continuation.resumeWithException(
                        Exception("HTTP ${response.code}: ${response.message}")
                    )
                }
            }
        })
    }

    data class ChatResponse(
        val response: String,
        val mode: String,
        val tokens: Int
    )
}

4. ViewModel 사용 예제

// ChatViewModel.kt
import androidx.lifecycle.ViewModel
import androidx.lifecycle.viewModelScope
import kotlinx.coroutines.flow.MutableStateFlow
import kotlinx.coroutines.flow.StateFlow
import kotlinx.coroutines.flow.catch
import kotlinx.coroutines.launch

class ChatViewModel : ViewModel() {

    private val apiClient = ChatBiaApiClient("https://your-hf-space.hf.space")

    private val _chatState = MutableStateFlow("")
    val chatState: StateFlow<String> = _chatState

    private val _isLoading = MutableStateFlow(false)
    val isLoading: StateFlow<Boolean> = _isLoading

    /**
     * 스트리밍 채팅 전송
     */
    fun sendStreamingMessage(message: String, mode: String = "bsl") {
        viewModelScope.launch {
            _isLoading.value = true
            _chatState.value = ""  // 초기화

            val request = ChatRequest(
                message = message,
                mode = mode,
                max_tokens = 1024,
                temperature = 0.7f
            )

            apiClient.chatStream(request)
                .catch { e ->
                    _chatState.value = "오류: ${e.message}"
                    _isLoading.value = false
                }
                .collect { response ->
                    if (response.error != null) {
                        _chatState.value = "서버 오류: ${response.error}"
                        _isLoading.value = false
                    } else if (response.done) {
                        // 완료
                        _isLoading.value = false
                    } else {
                        // 토큰 추가
                        _chatState.value += response.token
                    }
                }
        }
    }
}

5. Compose UI 예제

// ChatScreen.kt
import androidx.compose.foundation.layout.*
import androidx.compose.material3.*
import androidx.compose.runtime.*
import androidx.compose.ui.Modifier
import androidx.compose.ui.unit.dp

@Composable
fun ChatScreen(viewModel: ChatViewModel = viewModel()) {

    val chatState by viewModel.chatState.collectAsState()
    val isLoading by viewModel.isLoading.collectAsState()
    var inputText by remember { mutableStateOf("") }

    Column(
        modifier = Modifier
            .fillMaxSize()
            .padding(16.dp)
    ) {
        // 채팅 출력
        Card(
            modifier = Modifier
                .fillMaxWidth()
                .weight(1f)
        ) {
            Text(
                text = chatState,
                modifier = Modifier.padding(16.dp)
            )
        }

        Spacer(modifier = Modifier.height(16.dp))

        // 입력 필드
        Row(
            modifier = Modifier.fillMaxWidth()
        ) {
            OutlinedTextField(
                value = inputText,
                onValueChange = { inputText = it },
                modifier = Modifier.weight(1f),
                placeholder = { Text("메시지 입력...") },
                enabled = !isLoading
            )

            Spacer(modifier = Modifier.width(8.dp))

            Button(
                onClick = {
                    if (inputText.isNotBlank()) {
                        viewModel.sendStreamingMessage(inputText)
                        inputText = ""
                    }
                },
                enabled = !isLoading
            ) {
                Text(if (isLoading) "전송 중..." else "전송")
            }
        }
    }
}

---

🧪 테스트 방법

1. 로컬 서버 실행

cd ChatBIA-Server
uvicorn main:app --host 0.0.0.0 --port 8000

2. Python 테스트

python test_streaming.py

3. 안드로이드 앱에서 연결

// 로컬 테스트 (에뮬레이터)
val apiClient = ChatBiaApiClient("http://10.0.2.2:8000")

// 실제 디바이스 (같은 네트워크)
val apiClient = ChatBiaApiClient("http://YOUR_IP:8000")

// Hugging Face Spaces (배포 후)
val apiClient = ChatBiaApiClient("https://your-space.hf.space")

---

📊 응답 형식

스트리밍 응답 (SSE)

data: {"token":"안녕","done":false,"token_count":1}

data: {"token":"하세요","done":false,"token_count":2}

data: {"token":"!","done":false,"token_count":3}

data: {"token":"","done":true,"token_count":3,"mode":"bsl"}

최종 응답

{
  "token": "",
  "done": true,
  "token_count": 150,
  "mode": "bsl"
}

오류 응답

{
  "error": "오류 메시지",
  "done": true
}

---

⚡ 성능 최적화 팁

1. 타임아웃 설정

   • Connect: 30초
   • Read: 60초 (스트리밍)
   • Write: 30초

2. 재연결 로직

   fun retryOnFailure(maxRetries: Int = 3) {
       var attempts = 0
       while (attempts < maxRetries) {
           try {
               chatStream(request).collect { }
               break
           } catch (e: Exception) {
               attempts++
               delay(2000 * attempts)  // 지수 백오프
           }
       }
   }
   

3. 메모리 관리

   • Flow를 사용하여 메모리 효율적으로 처리
   • UI 업데이트는 StateFlow로 최적화

---

🚀 배포 후 사용

Hugging Face Spaces에 배포 후:

val BASE_URL = "https://your-username-chatbia-server.hf.space"
val apiClient = ChatBiaApiClient(BASE_URL)

주의: Hugging Face Spaces 무료 플랜은 CPU만 제공되므로 응답 속도가 느릴 수 있습니다. 스트리밍 방식이 더욱 중요합니다!

---

🔗 관련 문서

• FastAPI Streaming
• OkHttp SSE
• Kotlin Flow