VoiceWake: drop remote ssh config and harden template parsing

This commit is contained in:
Peter Steinberger
2025-12-09 03:04:08 +00:00
parent 4eb71bcd14
commit 2756e12762
6 changed files with 57 additions and 92 deletions

View File

@@ -88,14 +88,6 @@ final class AppState: ObservableObject {
didSet { UserDefaults.standard.set(self.voiceWakeForwardEnabled, forKey: voiceWakeForwardEnabledKey) }
}
@Published var voiceWakeForwardTarget: String {
didSet { UserDefaults.standard.set(self.voiceWakeForwardTarget, forKey: voiceWakeForwardTargetKey) }
}
@Published var voiceWakeForwardIdentity: String {
didSet { UserDefaults.standard.set(self.voiceWakeForwardIdentity, forKey: voiceWakeForwardIdentityKey) }
}
@Published var voiceWakeForwardCommand: String {
didSet { UserDefaults.standard.set(self.voiceWakeForwardCommand, forKey: voiceWakeForwardCommandKey) }
}
@@ -172,11 +164,6 @@ final class AppState: ObservableObject {
self.voiceWakeAdditionalLocaleIDs = UserDefaults.standard
.stringArray(forKey: voiceWakeAdditionalLocalesKey) ?? []
self.voiceWakeForwardEnabled = UserDefaults.standard.bool(forKey: voiceWakeForwardEnabledKey)
let legacyTarget = Self.legacyTargetString()
self.voiceWakeForwardTarget = UserDefaults.standard
.string(forKey: voiceWakeForwardTargetKey) ?? legacyTarget
self.voiceWakeForwardIdentity = UserDefaults.standard.string(forKey: voiceWakeForwardIdentityKey) ?? ""
self.voicePushToTalkEnabled = UserDefaults.standard
.object(forKey: voicePushToTalkEnabledKey) as? Bool ?? false
@@ -317,21 +304,9 @@ extension AppState {
var voiceWakeForwardConfig: VoiceWakeForwardConfig {
VoiceWakeForwardConfig(
enabled: self.voiceWakeForwardEnabled,
target: self.voiceWakeForwardTarget,
identityPath: self.voiceWakeForwardIdentity,
commandTemplate: self.voiceWakeForwardCommand,
timeout: defaultVoiceWakeForwardTimeout)
}
private static func legacyTargetString() -> String {
let host = UserDefaults.standard.string(forKey: voiceWakeForwardHostKey) ?? ""
let user = UserDefaults.standard.string(forKey: voiceWakeForwardUserKey) ?? ""
let savedPort = UserDefaults.standard.integer(forKey: voiceWakeForwardPortKey)
let port = savedPort == 0 ? defaultVoiceWakeForwardPort : savedPort
let userPrefix = user.isEmpty ? "" : "\(user)@"
let portSuffix = host.isEmpty ? "" : ":\(port)"
return "\(userPrefix)\(host)\(portSuffix)"
}
}
@MainActor

View File

@@ -16,11 +16,6 @@ let voiceWakeMicKey = "clawdis.voiceWakeMicID"
let voiceWakeLocaleKey = "clawdis.voiceWakeLocaleID"
let voiceWakeAdditionalLocalesKey = "clawdis.voiceWakeAdditionalLocaleIDs"
let voiceWakeForwardEnabledKey = "clawdis.voiceWakeForwardEnabled"
let voiceWakeForwardTargetKey = "clawdis.voiceWakeForwardTarget"
let voiceWakeForwardHostKey = "clawdis.voiceWakeForwardHost"
let voiceWakeForwardUserKey = "clawdis.voiceWakeForwardUser"
let voiceWakeForwardPortKey = "clawdis.voiceWakeForwardPort"
let voiceWakeForwardIdentityKey = "clawdis.voiceWakeForwardIdentity"
let voiceWakeForwardCommandKey = "clawdis.voiceWakeForwardCommand"
let voicePushToTalkEnabledKey = "clawdis.voicePushToTalkEnabled"
let iconOverrideKey = "clawdis.iconOverride"
@@ -36,6 +31,5 @@ let heartbeatsEnabledKey = "clawdis.heartbeatsEnabled"
let voiceWakeSupported: Bool = ProcessInfo.processInfo.operatingSystemVersion.majorVersion >= 26
let cliHelperSearchPaths = ["/usr/local/bin", "/opt/homebrew/bin"]
let defaultVoiceWakeForwardCommand = "clawdis-mac agent --message \"${text}\" --thinking low --session main --deliver"
let defaultVoiceWakeForwardPort = 22
// Allow enough time for remote agent responses (LLM replies often take >10s).
let defaultVoiceWakeForwardTimeout: TimeInterval = 30

View File

@@ -501,7 +501,11 @@ enum CommandResolver {
}
private static func sanitizedTarget(_ raw: String) -> String {
VoiceWakeForwarder.sanitizedTarget(raw)
let trimmed = raw.trimmingCharacters(in: .whitespacesAndNewlines)
if trimmed.hasPrefix("ssh ") {
return trimmed.replacingOccurrences(of: "ssh ", with: "").trimmingCharacters(in: .whitespacesAndNewlines)
}
return trimmed
}
private static func shellQuote(_ text: String) -> String {

View File

@@ -3,8 +3,6 @@ import OSLog
struct VoiceWakeForwardConfig: Sendable {
let enabled: Bool
let target: String
let identityPath: String
let commandTemplate: String
let timeout: TimeInterval
}
@@ -30,13 +28,11 @@ enum VoiceWakeForwarder {
}
enum VoiceWakeForwardError: LocalizedError, Equatable {
case invalidTarget
case rpcFailed(String)
case disabled
var errorDescription: String? {
switch self {
case .invalidTarget: return "Missing or invalid target"
case let .rpcFailed(message): return message
case .disabled: return "Voice wake forwarding disabled"
}
@@ -72,7 +68,6 @@ enum VoiceWakeForwarder {
static func checkConnection(config: VoiceWakeForwardConfig) async -> Result<Void, VoiceWakeForwardError> {
guard config.enabled else { return .failure(.disabled) }
guard !self.sanitizedTarget(config.target).isEmpty else { return .failure(.invalidTarget) }
let status = await AgentRPC.shared.status()
if status.ok { return .success(()) }
return .failure(.rpcFailed(status.error ?? "agent rpc unreachable"))
@@ -84,42 +79,6 @@ enum VoiceWakeForwarder {
return "'\(replaced)'"
}
static func parse(target: String) -> (user: String?, host: String, port: Int)? {
guard !target.isEmpty else { return nil }
var remainder = target
if remainder.hasPrefix("ssh ") {
remainder = remainder.replacingOccurrences(of: "ssh ", with: "")
}
remainder = remainder.trimmingCharacters(in: .whitespacesAndNewlines)
var user: String?
if let at = remainder.firstIndex(of: "@") {
user = String(remainder[..<at])
remainder = String(remainder[remainder.index(after: at)...])
}
var host = remainder
var port = defaultVoiceWakeForwardPort
if let colon = remainder.lastIndex(of: ":"), colon != remainder.startIndex {
let p = String(remainder[remainder.index(after: colon)...])
if let parsedPort = Int(p) {
port = parsedPort
host = String(remainder[..<colon])
}
}
host = host.trimmingCharacters(in: .whitespacesAndNewlines)
guard !host.isEmpty else { return nil }
return (user: user?.trimmingCharacters(in: .whitespacesAndNewlines), host: host, port: port)
}
static func sanitizedTarget(_ raw: String) -> String {
let trimmed = raw.trimmingCharacters(in: .whitespacesAndNewlines)
if trimmed.hasPrefix("ssh ") {
return trimmed.replacingOccurrences(of: "ssh ", with: "").trimmingCharacters(in: .whitespacesAndNewlines)
}
return trimmed
}
// MARK: - Template parsing
struct ForwardOptions {
@@ -131,7 +90,7 @@ enum VoiceWakeForwarder {
private static func parseCommandTemplate(_ template: String) -> ForwardOptions {
var options = ForwardOptions()
let parts = template.split(whereSeparator: { $0.isWhitespace }).map(String.init)
let parts = self.tokenize(template)
var idx = 0
while idx < parts.count {
let part = parts[idx]
@@ -157,6 +116,49 @@ enum VoiceWakeForwarder {
return options
}
private static func tokenize(_ text: String) -> [String] {
var tokens: [String] = []
var current = ""
var quote: Character?
var escapeNext = false
func flush() {
if !current.isEmpty {
tokens.append(current)
current = ""
}
}
for ch in text {
if escapeNext {
current.append(ch)
escapeNext = false
continue
}
if ch == "\\" && quote == "\"" {
escapeNext = true
continue
}
if ch == "\"" || ch == "'" {
if quote == ch {
quote = nil
} else if quote == nil {
quote = ch
} else {
current.append(ch)
}
continue
}
if ch.isWhitespace && quote == nil {
flush()
continue
}
current.append(ch)
}
flush()
return tokens
}
#if DEBUG
static func _testParseCommandTemplate(_ template: String) -> ForwardOptions {
self.parseCommandTemplate(template)

View File

@@ -125,7 +125,7 @@ final class VoiceWakeOverlayController: ObservableObject {
self.model.isSending = true
let payload = VoiceWakeForwarder.prefixedTranscript(text)
self.logger.log(level: .info, "overlay sendNow forwarding len=\(payload.count, privacy: .public) target=\(forwardConfig.target, privacy: .public)")
self.logger.log(level: .info, "overlay sendNow forwarding len=\(payload.count, privacy: .public)")
Task.detached {
await VoiceWakeForwarder.forward(transcript: payload, config: forwardConfig)
}