Gemini Configuration

Set up Google Gemini API for timeline analysis

Gemini offers the most efficient processing pipeline with native video understanding, requiring only 2 LLM calls per batch.

Getting a Gemini API Key

1. Visit Google AI Studio

   Go to https://ai.google.dev/gemini-api/docs/api-key

2. Create or select a project

   Sign in with your Google account and create a new project or select an existing one

3. Get your API key

   Click “Get API key” and copy the key. Keep it secure - anyone with this key can use your Gemini quota.

Do not commit your API key to version control or share it publicly. Treat it like a password.

Setting Up API Key in Dayflow

1. Open Settings

   Launch Dayflow and go to Settings → AI Provider

2. Select Gemini

   Choose Gemini from the provider options

3. Enter API Key

   Paste your API key into the field and save

4. Test the connection

   Start recording and verify that batches are processed successfully

Paid Services Data Handling

By default, Google may use your API data to improve their products. To prevent this:

1. Enable Cloud Billing

   In Google Cloud Console, enable Cloud Billing on at least one Gemini API project

2. Verify Paid Status

   Go to AI Studio → API keys and verify you see “Plan: Paid”

3. Data handling changes

   Once billing is enabled, all Gemini API usage (including free quota) is treated as “Paid Services” - Google will not use your prompts/responses to improve products

EEA/UK/Switzerland users: Paid-style data handling applies by default to all services, even without enabling billing.

Source: Gemini API Additional Terms - Paid Services

What Google Still Logs

Even under Paid Services, Google logs prompts/responses for a limited period for:

• Policy enforcement
• Legal compliance
• Abuse monitoring

This data is not used for training but is temporarily retained for safety purposes.

Model Selection

Dayflow supports multiple Gemini models with automatic fallback:

Available Models

Model	Display Name	Use Case	Fallback
gemini-3-flash-preview	Gemini 3 Flash (Default)	Best quality	→ 2.5 Flash → Flash Lite
gemini-2.5-flash	Gemini 2.5 Flash	Balanced	→ Flash Lite
gemini-2.5-flash-lite	Flash Lite	Most affordable	None

Fallback Behavior

If the selected model encounters capacity errors (429, 503, 403), Dayflow automatically falls back to the next model in the chain:

GeminiDirectProvider.swift:466-477

private func fallbackReason(for code: Int) -> String {
    switch code {
    case 429:
        return "rate_limit_429"
    case 503:
        return "service_unavailable_503"
    case 403:
        return "forbidden_quota_403"
    default:
        return "http_\(code)"
    }
}

Fallback switching is logged in the app and reported to analytics so you can monitor model availability.

Pricing

Gemini API offers generous free quota:

• 15 requests per minute (free tier)
• 1,500 requests per day (free tier)
• 1 million tokens per month (free tier)

For current pricing details, visit: https://ai.google.dev/gemini-api/docs/pricing

Processing Pipeline

Gemini’s efficiency comes from native video understanding:

Total: 2 LLM calls per 15-minute batch

Upload Flow

Dayflow uses resumable uploads for video chunks:

GeminiDirectProvider.swift:953-1032

private func uploadResumable(data: Data, mimeType: String) async throws -> String {
    print("📤 Starting resumable video upload:")
    print("   Size: \(data.count / 1024 / 1024) MB")
    print("   MIME Type: \(mimeType)")
    
    // 1. Initialize upload
    var request = URLRequest(url: URL(string: fileEndpoint + "?key=\(apiKey)")!)
    request.httpMethod = "POST"
    request.setValue("resumable", forHTTPHeaderField: "X-Goog-Upload-Protocol")
    request.setValue("start", forHTTPHeaderField: "X-Goog-Upload-Command")
    request.setValue("\(data.count)", forHTTPHeaderField: "X-Goog-Upload-Raw-Size")
    
    // 2. Get upload URL
    let (responseData, response) = try await URLSession.shared.data(for: request)
    guard let uploadURL = (response as? HTTPURLResponse)?.value(forHTTPHeaderField: "X-Goog-Upload-URL") else {
        throw NSError(domain: "GeminiError", code: 4, userInfo: [...])
    }
    
    // 3. Upload video data
    var uploadRequest = URLRequest(url: URL(string: uploadURL)!)
    uploadRequest.httpMethod = "PUT"
    uploadRequest.setValue("upload, finalize", forHTTPHeaderField: "X-Goog-Upload-Command")
    uploadRequest.httpBody = data
    
    let (uploadResponseData, _) = try await URLSession.shared.data(for: uploadRequest)
    // Parse and return file URI
}

Troubleshooting

Invalid API Key

If you see “Invalid API key” errors:

1. Verify the key is correct in Settings
2. Check that the API key has Gemini API access enabled
3. Ensure there are no extra spaces when pasting

Rate Limits

If you hit rate limits (429 errors):

1. Dayflow will automatically retry with exponential backoff
2. Consider upgrading to a paid plan for higher quotas
3. Check your quota usage in Google Cloud Console

Video Upload Failures

Dayflow implements retry logic with multiple cycles:

GeminiDirectProvider.swift:826-904

private func uploadAndAwait(_ fileURL: URL, mimeType: String, key: String, maxWaitTime: TimeInterval = 3 * 60) async throws -> (fileSize: Int64, fileURI: String) {
    let fileData = try Data(contentsOf: fileURL)
    let fileSize = fileData.count

    // Full cycle retry: upload + processing
    let maxCycles = 3
    var lastError: Error?

    for cycle in 1...maxCycles {
        print("🔄 Upload+Processing cycle \(cycle)/\(maxCycles)")

        // Upload with retries (3 attempts)
        var uploadedFileURI: String? = nil
        let maxUploadRetries = 3
        
        // ... upload retry logic ...
        
        // Poll for processing with 3-minute timeout
        let startTime = Date()
        while Date().timeIntervalSince(startTime) < maxWaitTime {
            let status = try await getFileStatus(fileURI: fileURI)
            if status == "ACTIVE" {
                return (Int64(fileSize), fileURI)
            }
            try await Task.sleep(nanoseconds: 2_000_000_000) // 2 seconds
        }
    }
}

If uploads consistently fail:

1. Check your internet connection
2. Verify the video file isn’t corrupted
3. Try restarting Dayflow