From fb34c46074b0e01807080d0abb946ec156cd47eb Mon Sep 17 00:00:00 2001 From: Ayaan Zaidi Date: Fri, 27 Feb 2026 08:57:29 +0530 Subject: [PATCH] refactor(android): make camera clip transport deterministic --- .../java/ai/openclaw/android/NodeRuntime.kt | 2 - .../ai/openclaw/android/node/CameraHandler.kt | 90 +++++-------------- 2 files changed, 22 insertions(+), 70 deletions(-) diff --git a/apps/android/app/src/main/java/ai/openclaw/android/NodeRuntime.kt b/apps/android/app/src/main/java/ai/openclaw/android/NodeRuntime.kt index 97a16d7af91..614cb957a2a 100644 --- a/apps/android/app/src/main/java/ai/openclaw/android/NodeRuntime.kt +++ b/apps/android/app/src/main/java/ai/openclaw/android/NodeRuntime.kt @@ -65,8 +65,6 @@ class NodeRuntime(context: Context) { private val cameraHandler: CameraHandler = CameraHandler( appContext = appContext, camera = camera, - prefs = prefs, - connectedEndpoint = { connectedEndpoint }, externalAudioCaptureActive = externalAudioCaptureActive, showCameraHud = ::showCameraHud, triggerCameraFlash = ::triggerCameraFlash, diff --git a/apps/android/app/src/main/java/ai/openclaw/android/node/CameraHandler.kt b/apps/android/app/src/main/java/ai/openclaw/android/node/CameraHandler.kt index f142a11f82e..ff1b8468cd6 100644 --- a/apps/android/app/src/main/java/ai/openclaw/android/node/CameraHandler.kt +++ b/apps/android/app/src/main/java/ai/openclaw/android/node/CameraHandler.kt @@ -3,8 +3,6 @@ package ai.openclaw.android.node import android.content.Context import ai.openclaw.android.CameraHudKind import ai.openclaw.android.BuildConfig -import ai.openclaw.android.SecurePrefs -import ai.openclaw.android.gateway.GatewayEndpoint import ai.openclaw.android.gateway.GatewaySession import kotlinx.coroutines.Dispatchers import kotlinx.coroutines.flow.MutableStateFlow @@ -12,27 +10,15 @@ import kotlinx.coroutines.withContext import kotlinx.serialization.json.Json import kotlinx.serialization.json.JsonPrimitive import kotlinx.serialization.json.contentOrNull -import okhttp3.MediaType.Companion.toMediaType -import okhttp3.RequestBody.Companion.asRequestBody -internal fun parseCameraClipUploadUrl(responseBody: String): String? { - if (responseBody.isBlank()) return null - val root = - try { - Json.parseToJsonElement(responseBody).asObjectOrNull() - } catch (_: Throwable) { - return null - } ?: return null - val urlPrimitive = root["url"] as? JsonPrimitive ?: return null - if (!urlPrimitive.isString) return null - return urlPrimitive.contentOrNull?.trim()?.ifEmpty { null } -} +internal const val CAMERA_CLIP_MAX_RAW_BYTES: Long = 18L * 1024L * 1024L + +internal fun isCameraClipWithinPayloadLimit(rawBytes: Long): Boolean = + rawBytes in 0L..CAMERA_CLIP_MAX_RAW_BYTES class CameraHandler( private val appContext: Context, private val camera: CameraCaptureManager, - private val prefs: SecurePrefs, - private val connectedEndpoint: () -> GatewayEndpoint?, private val externalAudioCaptureActive: MutableStateFlow, private val showCameraHud: (message: String, kind: CameraHudKind, autoHideMs: Long?) -> Unit, private val triggerCameraFlash: () -> Unit, @@ -105,60 +91,28 @@ class CameraHandler( showCameraHud(message, CameraHudKind.Error, 2400) return GatewaySession.InvokeResult.error(code = code, message = message) } - // Upload file via HTTP instead of base64 through WebSocket - clipLog("uploading via HTTP...") - val uploadUrl = try { - withContext(Dispatchers.IO) { - val ep = connectedEndpoint() - val gatewayHost = if (ep != null) { - val isHttps = ep.tlsEnabled || ep.port == 443 - if (!isHttps) { - clipLog("refusing to upload over plain HTTP — bearer token would be exposed; falling back to base64") - throw Exception("HTTPS required for upload (bearer token protection)") - } - if (ep.port == 443) "https://${ep.host}" else "https://${ep.host}:${ep.port}" - } else { - clipLog("error: no gateway endpoint connected, cannot upload") - throw Exception("no gateway endpoint connected") - } - val token = prefs.loadGatewayToken() ?: "" - val client = okhttp3.OkHttpClient.Builder() - .connectTimeout(10, java.util.concurrent.TimeUnit.SECONDS) - .writeTimeout(120, java.util.concurrent.TimeUnit.SECONDS) - .readTimeout(30, java.util.concurrent.TimeUnit.SECONDS) - .build() - val body = filePayload.file.asRequestBody("video/mp4".toMediaType()) - val req = okhttp3.Request.Builder() - .url("$gatewayHost/upload/clip.mp4") - .put(body) - .header("Authorization", "Bearer $token") - .build() - clipLog("uploading ${filePayload.file.length()} bytes to $gatewayHost/upload/clip.mp4") - val resp = client.newCall(req).execute() - val respBody = resp.body?.string() ?: "" - clipLog("upload response: ${resp.code} $respBody") - filePayload.file.delete() - if (!resp.isSuccessful) throw Exception("upload failed: HTTP ${resp.code}") - parseCameraClipUploadUrl(respBody) ?: throw Exception("no url in response: $respBody") - } - } catch (err: Throwable) { - clipLog("upload failed: ${err.message}, falling back to base64") - // Fallback to base64 if upload fails - val bytes = withContext(Dispatchers.IO) { - val b = filePayload.file.readBytes() - filePayload.file.delete() - b - } - val base64 = android.util.Base64.encodeToString(bytes, android.util.Base64.NO_WRAP) - showCameraHud("Clip captured", CameraHudKind.Success, 1800) - return GatewaySession.InvokeResult.ok( - """{"format":"mp4","base64":"$base64","durationMs":${filePayload.durationMs},"hasAudio":${filePayload.hasAudio}}""" + val rawBytes = filePayload.file.length() + if (!isCameraClipWithinPayloadLimit(rawBytes)) { + clipLog("payload too large: bytes=$rawBytes max=$CAMERA_CLIP_MAX_RAW_BYTES") + withContext(Dispatchers.IO) { filePayload.file.delete() } + showCameraHud("Clip too large", CameraHudKind.Error, 2400) + return GatewaySession.InvokeResult.error( + code = "PAYLOAD_TOO_LARGE", + message = + "PAYLOAD_TOO_LARGE: camera clip is $rawBytes bytes; max is $CAMERA_CLIP_MAX_RAW_BYTES bytes. Reduce durationMs and retry.", ) } - clipLog("returning URL result: $uploadUrl") + + val bytes = withContext(Dispatchers.IO) { + val b = filePayload.file.readBytes() + filePayload.file.delete() + b + } + val base64 = android.util.Base64.encodeToString(bytes, android.util.Base64.NO_WRAP) + clipLog("returning base64 payload") showCameraHud("Clip captured", CameraHudKind.Success, 1800) return GatewaySession.InvokeResult.ok( - """{"format":"mp4","url":"$uploadUrl","durationMs":${filePayload.durationMs},"hasAudio":${filePayload.hasAudio}}""" + """{"format":"mp4","base64":"$base64","durationMs":${filePayload.durationMs},"hasAudio":${filePayload.hasAudio}}""" ) } catch (err: Throwable) { clipLog("outer error: ${err::class.java.simpleName}: ${err.message}")