Skip to main content

​
What is Groq?

Groq powers the AI features in Reflect AI, including:
  • Personalized greetings
  • Weekly and monthly summaries
  • Writing suggestions
  • Entry rewriting
  • Entry generation from quick notes

​
Privacy-First AI

AI features are entirely optional. Reflect AI works fully without a Groq API key. Sentiment analysis runs locally using NLTK/VADER.

​
What Gets Sent to Groq?

When you use AI features, only the specific content needed is sent:

​
Greeting Generation

  • Your total entry count
  • Current streak
  • Recent mood summary (e.g., “mostly positive”)
  • Recent themes (e.g., “work, health”)
  • NOT sent: Full entry text

​
Weekly/Monthly Summaries

  • Preview of recent entries (first 100-300 characters)
  • Detected mood labels
  • Detected themes and activities
  • Pattern insights
  • NOT sent: Your complete entry text

​
Writing Features

  • The specific text you want to rewrite/expand
  • Context from recent entry previews (optional)
  • NOT sent: Your entire journal history

​
What Groq Does NOT Store

According to Groq’s API terms, prompts sent via API are not used to train models and are not retained after processing.
Groq processes your data to generate responses, then discards it. They do not: ❌ Store your journal entries
❌ Use your data for AI training
❌ Share your data with third parties
❌ Build a profile on you However, you should review Groq’s privacy policy for the most current information.

​
API Key Setup

Your Groq API key is stored in a .env file on your device: 
GROQ_API_KEY=your_api_key_here

​
Getting a Free API Key

  1. Visit console.groq.com
  2. Create a free account
  3. Generate an API key
  4. Add it to your .env file
Groq offers a generous free tier. No credit card required.

​
How AI Features Work

Here’s the data flow:

​
Data Minimization

Reflect AI follows a data minimization principle when using AI:
  1. Only send what’s needed: Previews, not full text
  2. Aggregate when possible: “mostly positive” instead of full sentiment data
  3. You control when: AI features are opt-in per use
  4. Fallback available: App works without API key

​
Model Used

Reflect AI uses the llama-3.1-8b-instant model via Groq.
This is a fast, open-source model that provides quality results for journaling features.

​
Security Recommendations

​
Protect Your API Key

Your API key provides access to your Groq account. Keep it secure.
  • Store in .env file (not tracked by git)
  • Don’t share your API key
  • Regenerate if compromised
  • Monitor usage on Groq dashboard

​
Disable AI Features

To completely disable AI features:
  1. Remove the GROQ_API_KEY from your .env file
  2. Restart the app
The app will fall back to local-only features:
  • Sentiment analysis (NLTK VADER - fully local)
  • Theme detection (keyword-based - fully local)
  • Activity correlation (pattern matching - fully local)
  • Stats and charts (computed locally)

​
Privacy Implications

Local-Only Features

No data leaves your device
  • Sentiment analysis
  • Theme detection
  • Streak tracking
  • Charts and stats

AI Features

Minimal data sent to Groq
  • Greetings (metadata only)
  • Summaries (previews)
  • Writing help (selected text)
  • Not retained after processing

​
Transparency

When an AI feature is used, the app indicates it in the response: 
{
  "message": "Your personalized greeting...",
  "ai_generated": true
}
You always know when AI was involved.

​
Questions?

​
Can Groq employees read my journal?

No. API requests are processed programmatically and not viewed by humans. Groq does not retain prompt data after processing.

​
What if Groq changes their policy?

Reflect AI is designed to work without AI features. You can disable Groq integration at any time and continue using the app.

​
Can I use a different AI provider?

Currently, only Groq is supported. However, the code is open-source and can be modified to use other providers (OpenAI, Anthropic, local models, etc.).

​
Does local AI work?

You could modify the code to use local models (Ollama, llamafile) for complete offline AI. This would require code changes but is technically possible.

​
Best Practices

  1. Review prompts: Check what data is sent before using AI features
  2. Use selectively: AI features are optional; use them when beneficial
  3. Trust but verify: Review AI-generated content before using it
  4. Stay informed: Monitor Groq’s privacy policy for changes
  5. Local when possible: Prefer local features for maximum privacy

​
Summary

Groq integration is privacy-conscious by design:
  • Optional, not required
  • Minimal data sent
  • Not retained after processing
  • Transparent when used
  • Full control to disable

Build docs developers (and LLMs) love

Get started for freeTalk to us