feat(polish): allow mood emoji on custom styles and ship Flow/ASR fixes

Custom polish styles can opt in to emotion-matched emoji (default off), with prompt-level opt-in detection so paste-only styles keep model-added emoji. Also include Volcengine API-Key ASR auth, voice-processing capture, PiP flash fix, and related keyboard Shift/haptics reliability work.
2026-08-06 15:11:12 +08:00
parent 2ed16e1fbf
commit a8d58d8f0c
47 changed files with 1467 additions and 258 deletions
@@ -140,7 +140,13 @@ public struct AppGroupConfiguration: Sendable, Equatable {

    public var isCloudASRKeyMissing: Bool {
        guard engineMode == "cloud" else { return false }
-        return asrApiKey.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
+        let key = asrApiKey.trimmingCharacters(in: .whitespacesAndNewlines)
+        guard !key.isEmpty else { return true }
+        // Volcengine may store auth_mode JSON before credentials are filled.
+        if asrProviderId == "volcengine" {
+            return !VolcengineASRFields.parse(apiKey: key).hasUsableCredentials
+        }
+        return false
    }

    public var isPolishKeyMissing: Bool {
@@ -191,7 +191,7 @@ public struct LLMProvider: Identifiable, Codable, Hashable, Sendable {
            defaultBaseURL: "wss://openspeech.bytedance.com/api/v3/sauc/bigmodel_async",
            defaultModel: "volc.seedasr.sauc.duration",
            apiKeyURL: URL(string: "https://console.volcengine.com/speech"),
-            blurb: "流式大模型 ASR · API Key 填 appId:accessToken[:resourceId]",
+            blurb: "流式大模型 ASR · 旧版 AppID+Token / 新版 API Key",
            isUserSelectable: false
        ),
        .init(
@@ -15,6 +15,9 @@ public struct PolishStylePack: Codable, Equatable, Identifiable, Sendable {
    public let id: String
    public var name: String
    public var prompt: String
+    /// When true, polish may keep model-added emoji and the prompt overrides R5.
+    /// Defaults off so existing / builtin styles stay emoji-strict.
+    public var allowsAddedEmoji: Bool
    public let kind: Kind
    public let createdAt: Date
    public var updatedAt: Date
@@ -23,6 +26,7 @@ public struct PolishStylePack: Codable, Equatable, Identifiable, Sendable {
        id: String = "user.\(UUID().uuidString.lowercased())",
        name: String,
        prompt: String,
+        allowsAddedEmoji: Bool = false,
        kind: Kind = .user,
        createdAt: Date = Date(),
        updatedAt: Date? = nil
@@ -30,6 +34,7 @@ public struct PolishStylePack: Codable, Equatable, Identifiable, Sendable {
        self.id = id
        self.name = name
        self.prompt = prompt
+        self.allowsAddedEmoji = allowsAddedEmoji
        self.kind = kind
        self.createdAt = createdAt
        self.updatedAt = updatedAt ?? createdAt
@@ -39,6 +44,46 @@ public struct PolishStylePack: Codable, Equatable, Identifiable, Sendable {
        guard kind == .builtin else { return name }
        return SharedL10n.string("polishStyle.\(id.dropFirst("builtin.".count))", language: language)
    }
+
+    /// Effective emoji policy for polish: explicit toggle, or a custom prompt that
+    /// clearly opts in (so paste-only custom styles still keep model-added emoji).
+    public var effectiveAllowsAddedEmoji: Bool {
+        if allowsAddedEmoji { return true }
+        guard kind == .user else { return false }
+        return Self.promptDeclaresAddedEmojiOptIn(prompt)
+    }
+
+    /// Heuristic for custom prompts that declare “add mood emoji” themselves.
+    public static func promptDeclaresAddedEmojiOptIn(_ prompt: String) -> Bool {
+        let markers = [
+            "允许新增 emoji",
+            "允许新增emoji",
+            "按情绪点缀",
+            "按原文情绪",
+            "Emoji 覆盖",
+            "outranks global R5",
+            "may add emojis",
+            "allow mood emoji",
+            "allowsAddedEmoji",
+        ]
+        return markers.contains { prompt.localizedCaseInsensitiveContains($0) }
+    }
+
+    private enum CodingKeys: String, CodingKey {
+        case id, name, prompt, allowsAddedEmoji, kind, createdAt, updatedAt
+    }
+
+    public init(from decoder: Decoder) throws {
+        let container = try decoder.container(keyedBy: CodingKeys.self)
+        id = try container.decode(String.self, forKey: .id)
+        name = try container.decode(String.self, forKey: .name)
+        prompt = try container.decode(String.self, forKey: .prompt)
+        // Older synced packs omit the key — stay emoji-strict.
+        allowsAddedEmoji = try container.decodeIfPresent(Bool.self, forKey: .allowsAddedEmoji) ?? false
+        kind = try container.decode(Kind.self, forKey: .kind)
+        createdAt = try container.decode(Date.self, forKey: .createdAt)
+        updatedAt = try container.decode(Date.self, forKey: .updatedAt)
+    }
 }

 public enum PolishStyleLimits {
@@ -92,6 +137,7 @@ public struct PolishStyleCatalog: Codable, Equatable, Sendable {
            var updated = pack
            updated.name = name
            updated.prompt = prompt
+            updated.allowsAddedEmoji = pack.allowsAddedEmoji
            updated.updatedAt = date
            entries[index] = updated
        } else {
@@ -101,6 +147,7 @@ public struct PolishStyleCatalog: Codable, Equatable, Sendable {
            var created = pack
            created.name = name
            created.prompt = prompt
+            created.allowsAddedEmoji = pack.allowsAddedEmoji
            created.updatedAt = date
            entries.append(created)
        }
@@ -313,7 +313,14 @@ public final class ProviderConfig: ObservableObject, @unchecked Sendable {

    public var isASRConfigured: Bool {
        guard !isLocalEngine else { return true }
-        return !asrApiKey.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
+        let key = asrApiKey.trimmingCharacters(in: .whitespacesAndNewlines)
+        let hasKey: Bool = {
+            if asrProviderId == "volcengine" {
+                return VolcengineASRFields.parse(apiKey: key).hasUsableCredentials
+            }
+            return !key.isEmpty
+        }()
+        return hasKey
            && (!asrBaseURL.isEmpty || CloudASRModelCatalog.strategy(for: asrProviderId) != .prompt)
    }

@@ -2,65 +2,155 @@
 // OSGKeyboard · Shared
 //
 // Parse / encode Volcengine SAUC credentials stored in the ASR API key field.
+// Supports legacy AppID+AccessToken (old console) and single API Key (new console).

 import Foundation

+/// Volcengine speech console auth style for SAUC streaming ASR.
+public enum VolcengineASRAuthMode: String, Sendable, Equatable {
+    /// Old console: `X-Api-App-Key` + `X-Api-Access-Key`.
+    case appToken = "app_token"
+    /// New console: single `X-Api-Key`.
+    case apiKey = "api_key"
+}
+
 public struct VolcengineASRFields: Sendable, Equatable {
+    public var authMode: VolcengineASRAuthMode
    public var appID: String
    public var accessToken: String
-    public var resourceID: String
+    /// New-console API Key (`X-Api-Key`). Kept alongside app-token fields so the
+    /// settings toggle can switch modes without wiping the other credential set.
+    public var apiKeyCredential: String
+
+    /// Locked product: Doubao streaming ASR 2.0 · duration billing.
+    public static let fixedResourceID = CloudASRModelCatalog.volcengineDefaultResourceID

    public init(
+        authMode: VolcengineASRAuthMode = .appToken,
        appID: String = "",
        accessToken: String = "",
-        resourceID: String = CloudASRModelCatalog.defaultModel(for: "volcengine")
+        apiKeyCredential: String = ""
    ) {
+        self.authMode = authMode
        self.appID = appID
        self.accessToken = accessToken
-        self.resourceID = resourceID
+        self.apiKeyCredential = apiKeyCredential
+    }
+
+    /// Always `volc.seedasr.sauc.duration` — not user-editable.
+    public var resourceID: String { Self.fixedResourceID }
+
+    public var usesAPIKeyAuth: Bool { authMode == .apiKey }
+
+    public var hasUsableCredentials: Bool {
+        switch authMode {
+        case .appToken:
+            return !appID.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
+                && !accessToken.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
+        case .apiKey:
+            return !apiKeyCredential.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
+        }
    }

    public var encodedAPIKey: String {
-        let object = [
-            "app_id": appID,
-            "access_token": accessToken,
-            "resource_id": resourceID,
+        var object: [String: String] = [
+            "auth_mode": authMode.rawValue,
+            "resource_id": Self.fixedResourceID,
        ]
+        // Persist both credential sets so toggling auth mode is non-destructive.
+        if !appID.isEmpty { object["app_id"] = appID }
+        if !accessToken.isEmpty { object["access_token"] = accessToken }
+        if !apiKeyCredential.isEmpty { object["api_key"] = apiKeyCredential }
+
        guard let data = try? JSONSerialization.data(withJSONObject: object),
              let string = String(data: data, encoding: .utf8) else {
-            return [appID, accessToken, resourceID].joined(separator: ":")
+            switch authMode {
+            case .appToken:
+                return [appID, accessToken, Self.fixedResourceID].joined(separator: ":")
+            case .apiKey:
+                return apiKeyCredential
+            }
        }
        return string
    }

-    public static func parse(apiKey: String, resourceFallback: String) -> VolcengineASRFields {
+    /// Apply SAUC WebSocket handshake headers for the active auth mode.
+    public func applyWebSocketAuthHeaders(to request: inout URLRequest, connectID: String) {
+        request.setValue(Self.fixedResourceID, forHTTPHeaderField: "X-Api-Resource-Id")
+        request.setValue(connectID, forHTTPHeaderField: "X-Api-Connect-Id")
+        switch authMode {
+        case .apiKey:
+            request.setValue(
+                apiKeyCredential.trimmingCharacters(in: .whitespacesAndNewlines),
+                forHTTPHeaderField: "X-Api-Key"
+            )
+        case .appToken:
+            request.setValue(
+                appID.trimmingCharacters(in: .whitespacesAndNewlines),
+                forHTTPHeaderField: "X-Api-App-Key"
+            )
+            request.setValue(
+                accessToken.trimmingCharacters(in: .whitespacesAndNewlines),
+                forHTTPHeaderField: "X-Api-Access-Key"
+            )
+        }
+    }
+
+    /// - Parameter resourceFallback: Ignored; resource is always `fixedResourceID`.
+    ///   Kept so call sites stay source-compatible.
+    public static func parse(apiKey: String, resourceFallback: String = "") -> VolcengineASRFields {
+        _ = resourceFallback
        let trimmed = apiKey.trimmingCharacters(in: .whitespacesAndNewlines)
-        var fields = VolcengineASRFields(
-            appID: "",
-            accessToken: "",
-            resourceID: resourceFallback.isEmpty
-                ? CloudASRModelCatalog.defaultModel(for: "volcengine")
-                : resourceFallback
-        )
+        var fields = VolcengineASRFields()
+
+        guard !trimmed.isEmpty else { return fields }

        if let data = trimmed.data(using: .utf8),
           let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any] {
            fields.appID = string(json, keys: ["app_id", "appId", "appid"]) ?? ""
            fields.accessToken = string(json, keys: ["access_token", "accessToken", "token"]) ?? ""
-            fields.resourceID = string(json, keys: ["resource_id", "resourceId", "resource"]) ?? fields.resourceID
+            fields.apiKeyCredential = string(json, keys: ["api_key", "apiKey"]) ?? ""
+            fields.authMode = resolveAuthMode(
+                raw: string(json, keys: ["auth_mode", "authMode"]),
+                hasAPIKey: !fields.apiKeyCredential.isEmpty,
+                hasAppToken: !fields.appID.isEmpty && !fields.accessToken.isEmpty
+            )
            return fields
        }

+        // Legacy colon form is always old-console app-token auth.
        let parts = trimmed
            .components(separatedBy: CharacterSet(charactersIn: ":\n,"))
            .map { $0.trimmingCharacters(in: .whitespacesAndNewlines) }
            .filter { !$0.isEmpty }
        if parts.indices.contains(0) { fields.appID = parts[0] }
        if parts.indices.contains(1) { fields.accessToken = parts[1] }
-        if parts.indices.contains(2) { fields.resourceID = parts[2] }
+        fields.authMode = .appToken
        return fields
    }

+    private static func resolveAuthMode(
+        raw: String?,
+        hasAPIKey: Bool,
+        hasAppToken: Bool
+    ) -> VolcengineASRAuthMode {
+        if let raw {
+            let normalized = raw.trimmingCharacters(in: .whitespacesAndNewlines).lowercased()
+            if normalized == VolcengineASRAuthMode.apiKey.rawValue || normalized == "apikey" {
+                return .apiKey
+            }
+            if normalized == VolcengineASRAuthMode.appToken.rawValue
+                || normalized == "apptoken"
+                || normalized == "app_id_token" {
+                return .appToken
+            }
+        }
+        // Legacy JSON without auth_mode: prefer app-token when present.
+        if hasAppToken { return .appToken }
+        if hasAPIKey { return .apiKey }
+        return .appToken
+    }
+
    private static func string(_ json: [String: Any], keys: [String]) -> String? {
        for key in keys {
            if let value = json[key] as? String {