feat: mirror analytics events to stdout and OSLog by vegerot · Pull Request #228 · JerryZLiu/Dayflow

vegerot · 2026-03-05T22:42:37Z

Ensures every AnalyticsService event is sent to PostHog, printed to stdout, and emitted via Apple Unified Logging.

Stack created with Sapling. Best reviewed with ReviewStack.

Copilot

Pull request overview

Adds local mirroring for analytics events and refactors LLM prompt/transcript/timeline parsing to shared utilities, while extending connection testing UI to be provider-aware.

Changes:

Mirror AnalyticsService events to stdout and Apple Unified Logging alongside PostHog.
Extract/centralize LLM JSON schemas, prompt templates, transcript decoding, and timeline validation utilities.
Update onboarding/settings UI to support provider-specific connection testing and more robust card layout sizing.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
Dayflow/Dayflow/Views/UI/Settings/SettingsProvidersTabView.swift	Passes explicit provider into the connection test view.
Dayflow/Dayflow/Views/UI/Settings/ProvidersSettingsViewModel.swift	Switches Gemini prompt override persistence to shared “Video” prompt preferences types.
Dayflow/Dayflow/Views/Onboarding/TestConnectionView.swift	Makes connection tests provider-aware and standardizes analytics properties.
Dayflow/Dayflow/Views/Onboarding/OnboardingLLMSelectionView.swift	Computes card sizing based on dynamic provider count.
Dayflow/Dayflow/System/AnalyticsService.swift	Adds stdout + OSLog mirroring for analytics captures.
Dayflow/Dayflow/Core/Analysis/TimeParsing.swift	Introduces shared LLM timestamp, transcript decoding, and timeline validation helpers.
Dayflow/Dayflow/Core/AI/LLMSchema.swift	Adds centralized JSON schema strings for LLM structured outputs.
Dayflow/Dayflow/Core/AI/GeminiPromptPreferences.swift	Renames prompt override types and adds shared prompt template helpers.
Dayflow/Dayflow/Core/AI/GeminiDirectProvider.swift	Replaces inline prompts/parsers with shared schemas/utilities and adds schema-constrained generation config.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-05T22:45:34Z

Dayflow/Dayflow/Core/AI/GeminiDirectProvider.swift

+    let transcriptionSchemaObject = try! JSONSerialization.jsonObject(
+      with: Data(LLMSchema.screenRecordingTranscriptionSchema.utf8))
    let generationConfig: [String: Any] = [
      "temperature": 0.3,
      "maxOutputTokens": 65536,
      "mediaResolution": "MEDIA_RESOLUTION_HIGH",
      "responseMimeType": "application/json",
-      "responseSchema": transcriptionSchema,
+      "responseJsonSchema": transcriptionSchemaObject,
    ]


try! JSONSerialization.jsonObject(...) will crash the app at runtime if the schema string is ever invalid. Prefer a non-crashing path (e.g., pre-parse once and fail gracefully/throw) so schema changes can’t take the entire app down.

Copilot · 2026-03-05T22:45:35Z

Dayflow/Dayflow/Core/AI/GeminiDirectProvider.swift

+    let activityCardsSchemaObject = try? JSONSerialization.jsonObject(
+      with: Data(LLMSchema.activityCardsSchema.utf8))
    let generationConfig: [String: Any] = [
      "temperature": 0.7,
      "maxOutputTokens": 8192,
+      "responseMimeType": "application/json",
+      "responseJsonSchema": activityCardsSchemaObject,
    ]



activityCardsSchemaObject is an Any?; storing it into [String: Any] can embed an Optional.none value, which is not a valid JSON object and can break request serialization downstream. If schema parsing fails, omit responseJsonSchema entirely or throw early so the request body never contains an optional.

Suggested change

let activityCardsSchemaObject = try? JSONSerialization.jsonObject(

with: Data(LLMSchema.activityCardsSchema.utf8))

let generationConfig: [String: Any] = [

"temperature": 0.7,

"maxOutputTokens": 8192,

"responseMimeType": "application/json",

"responseJsonSchema": activityCardsSchemaObject,

]

var generationConfig: [String: Any] = [

"temperature": 0.7,

"maxOutputTokens": 8192,

"responseMimeType": "application/json",

]

if let activityCardsSchemaObject = try? JSONSerialization.jsonObject(

with: Data(LLMSchema.activityCardsSchema.utf8))

{

generationConfig["responseJsonSchema"] = activityCardsSchemaObject

}

Copilot · 2026-03-05T22:45:35Z

Dayflow/Dayflow/System/AnalyticsService.swift

+    let json = jsonString(properties)
+    let line = truncate("[Analytics] \(event) \(json)")
+    print(line)
+    localLogger.info("\(line, privacy: .public)")


Logging full analytics properties as .public (and printing to stdout) can expose potentially sensitive user data in Console logs / sysdiagnose artifacts. Consider making the OSLog message .private (or logging only the event name + a redacted subset), and/or gating the verbose property logging behind #if DEBUG or a dedicated runtime flag.

Suggested change

let json = jsonString(properties)

let line = truncate("[Analytics] \(event) \(json)")

print(line)

localLogger.info("\(line, privacy: .public)")

#if DEBUG

// In debug builds, log full analytics payload locally for diagnostics.

let json = jsonString(properties)

let line = truncate("[Analytics] \(event) \(json)")

print(line)

localLogger.info("\(line, privacy: .private)")

#else

// In non-debug builds, avoid logging full analytics properties.

localLogger.info("[Analytics] \(event, privacy: .public)")

#endif

Copilot · 2026-03-05T22:45:35Z

Dayflow/Dayflow/Core/Analysis/TimeParsing.swift

+  /// Returns `0` for invalid input.
+  static func parseVideoTimestamp(_ timestamp: String) -> Int {
+    let parts =
+      timestamp
+      .trimmingCharacters(in: .whitespacesAndNewlines)
+      .components(separatedBy: ":")
+
+    if parts.count == 3 {
+      let h = Int(parts[0]) ?? 0
+      let m = Int(parts[1]) ?? 0
+      let s = Int(parts[2]) ?? 0
+      return h * 3600 + m * 60 + s
+    }
+    if parts.count == 2 {
+      let m = Int(parts[0]) ?? 0
+      let s = Int(parts[1]) ?? 0
+      return m * 60 + s
+    }
+    return 0


Returning 0 for invalid timestamps makes invalid input indistinguishable from a legitimate 00:00 value. This can silently produce incorrect observation times and avoid incrementing invalidTimestampCount. Consider returning Int? (or throwing) and treating unparseable timestamps as invalid in LLMTranscriptUtilities.observations(...) and timeline validation.

Suggested change

/// Returns `0` for invalid input.

static func parseVideoTimestamp(_ timestamp: String) -> Int {

let parts =

timestamp

.trimmingCharacters(in: .whitespacesAndNewlines)

.components(separatedBy: ":")

if parts.count == 3 {

let h = Int(parts[0]) ?? 0

let m = Int(parts[1]) ?? 0

let s = Int(parts[2]) ?? 0

return h * 3600 + m * 60 + s

}

if parts.count == 2 {

let m = Int(parts[0]) ?? 0

let s = Int(parts[1]) ?? 0

return m * 60 + s

}

return 0

/// Returns `nil` for invalid or unparseable input.

static func parseVideoTimestamp(_ timestamp: String) -> Int? {

let trimmed = timestamp.trimmingCharacters(in: .whitespacesAndNewlines)

guard !trimmed.isEmpty else { return nil }

let parts = trimmed.components(separatedBy: ":")

switch parts.count {

case 3:

// HH:MM:SS – minutes/seconds must be in 0..<60, hours can be any non-negative integer.

guard

let h = Int(parts[0]), h >= 0,

let m = Int(parts[1]), (0..<60).contains(m),

let s = Int(parts[2]), (0..<60).contains(s)

else {

return nil

}

return h * 3600 + m * 60 + s

case 2:

// MM:SS – minutes can be any non-negative integer, seconds must be in 0..<60.

guard

let m = Int(parts[0]), m >= 0,

let s = Int(parts[1]), (0..<60).contains(s)

else {

return nil

}

return m * 60 + s

default:

return nil

}

Copilot · 2026-03-05T22:45:35Z

Dayflow/Dayflow/Core/Analysis/TimeParsing.swift

+    } catch {
+      let snippet = String(output.prefix(400))
+      print(
+        "🔎 LLM DEBUG: decodeTranscriptChunks JSON decode failed: \(error.localizedDescription) snippet=\(snippet)"
+      )


Printing a raw snippet of the model output can leak user/activity content into logs in production. Prefer OSLog with privacy redaction and/or restrict this verbose output to #if DEBUG.

Suggested change

} catch {

let snippet = String(output.prefix(400))

print(

"🔎 LLM DEBUG: decodeTranscriptChunks JSON decode failed: \(error.localizedDescription) snippet=\(snippet)"

)

} catch {

#if DEBUG

let snippet = String(output.prefix(400))

print(

"🔎 LLM DEBUG: decodeTranscriptChunks JSON decode failed: \(error.localizedDescription) snippet=\(snippet)"

)

#endif

Copilot · 2026-03-05T22:45:36Z

Dayflow/Dayflow/Core/AI/GeminiPromptPreferences.swift


-enum GeminiPromptPreferences {
+enum VideoPromptPreferences {
  private static let overridesKey = "geminiPromptOverrides"


VideoPromptPreferences still persists under the geminiPromptOverrides key, which is surprising after the rename and makes future maintenance/migrations harder. If the old key is intentional for backward compatibility, add an explicit comment and/or implement a one-time migration to a new key (while still reading the old key).

- Extract shared prompt templates into LLMPromptTemplates (GeminiPromptPreferences.swift) - Add VideoPromptPreferences/VideoPromptOverrides/VideoPromptSections types, replacing GeminiPromptPreferences/GeminiPromptOverrides/GeminiPromptSections - Centralize transcript JSON decoding and observation conversion in LLMTranscriptUtilities (TimeParsing.swift) for reuse across providers - Refactor GeminiDirectProvider to use LLMPromptTemplates and LLMTranscriptUtilities - Refactor TestConnectionView to accept a provider parameter with finishFailure/finishSuccess helpers for clean multi-provider support - Fix OnboardingLLMSelectionView card-width calculation to be dynamic based on card count rather than hard-coded divisor of 3 - Update SettingsProvidersTabView and ProvidersSettingsViewModel to use new VideoPrompt* types Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Ensures every AnalyticsService event is sent to PostHog, printed to stdout, and emitted via Apple Unified Logging.

Copilot AI review requested due to automatic review settings March 5, 2026 22:42

This was referenced Mar 5, 2026

feat: add Doubao (Ark) LLM provider #212

Open

refactor: centralize LLM prompt templates and transcript utilities #226

Open

feat: use structured output to improve responses #227

Open

Copilot AI reviewed Mar 5, 2026

View reviewed changes

Copilot started reviewing on behalf of vegerot March 5, 2026 23:02 View session

Copilot started reviewing on behalf of vegerot March 5, 2026 23:03 View session

vegerot force-pushed the pr228 branch from a8175f9 to ec9d571 Compare March 7, 2026 00:07

vegerot and others added 3 commits March 6, 2026 16:12

feat: use structured output to improve responses

dd71b08

feat: mirror analytics events to stdout and OSLog

ac550e8

Ensures every AnalyticsService event is sent to PostHog, printed to stdout, and emitted via Apple Unified Logging.

vegerot force-pushed the pr228 branch from ec9d571 to ac550e8 Compare March 7, 2026 00:12

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: mirror analytics events to stdout and OSLog#228

feat: mirror analytics events to stdout and OSLog#228
vegerot wants to merge 3 commits intoJerryZLiu:mainfrom
vegerot:pr228

vegerot commented Mar 5, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 5, 2026

Uh oh!

Copilot AI Mar 5, 2026

Uh oh!

Copilot AI Mar 5, 2026

Uh oh!

Copilot AI Mar 5, 2026

Uh oh!

Copilot AI Mar 5, 2026

Uh oh!

Copilot AI Mar 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

vegerot commented Mar 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 5, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 5, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

vegerot commented Mar 5, 2026 •

edited

Loading