What is Log Anything?
Log Anything is the AI-powered universal input on your Peanuts home screen. Using the Unified Input bar, just describe what you want to log — via text, voice, image, or link — and Peanuts automatically routes it to the right Helper.Log Anything uses a 4-tier matching system. Simple entries often log for 0 credits using Affinity or Local matching. Only ambiguous inputs require AI (1 credit).
The 4-Tier Matching Pipeline
When you type in the Unified Input, Peanuts processes your entry through four matching tiers:| Tier | Name | Credits | How It Works |
|---|---|---|---|
| 1 | Affinity | 0 ⚡ | Matches based on your learned patterns and history |
| 2 | Local | 0 ⚡ | Pattern matching against Helper field definitions |
| 3 | AI | 1 | OpenAI analysis for complex/ambiguous inputs |
| 4 | Fallback | 0 | Goes to Shoebox if no match |
Instant Logs ⚡
For simple, unambiguous inputs, Peanuts can log instantly without using AI credits.Instant ⚡ (0 credits)
- “coffee $5”
- “ran 3km”
- “slept 7 hours”
Requires AI (1 credit)
- “about $5” (ambiguous)
- “yesterday’s lunch” (time context)
- “total for the week” (calculation)
How to Use
Type directly in the Unified Input Bar at the bottom of your home screen. As you type, the Context Cloud appears showing:- Which Helper matches your input
- Which fields will be populated
- Confidence level of the match
Text
Type naturally: “Spent $45 at Whole Foods on groceries”
Voice
Tap the mic and speak your entry
Camera
Take a photo of a receipt, meal, document, or anything
Link
Paste a URL to save articles, recipes, products, or web content
How It Works
Fallback to Shoebox
If no confident match, entry goes to Shoebox (0 credits)
Clarification Pills
When your input could match multiple Helpers, Peanuts shows Clarification Pills — tappable options to disambiguate:Input Types
Text Input
Just type naturally:- “Coffee at Starbucks $5.50”
- “Ran 5k in 28 minutes”
- “Feeling tired today, slept badly”
- “Book recommendation: Atomic Habits”
Voice Input
Tap the microphone and speak:- Works in any language Peanuts supports
- Transcription is done instantly
- Same routing as text input
- Costs 9 credits (8 for transcription + 1 for routing)
Image Input
Take a photo or upload from gallery:- Receipts → Extracted and added to expense tracker
- Food → Analyzed and added to meal log
- Documents → Text extracted and stored
- Products → Identified and cataloged
- Anything else → AI figures out the best destination
Link Input
Paste any URL:- Articles → Title, summary, and content saved
- Recipes → Ingredients and instructions extracted
- Products → Name, price, and details captured
- Videos → Title and description saved
- Any webpage → Relevant content extracted
Shoebox Fallback
If Peanuts can’t confidently match your input to any Helper:- Entry goes to the Shoebox (0 credits)
- You can organize it later
- If similar items accumulate, Peanuts suggests creating a new Helper
Examples
Expense Tracking
Expense Tracking
Input: “Uber to airport $35”Result: Entry added to Expense Tracker with:
- Amount: $35
- Category: Transportation
- Description: Uber to airport
- Date: Today
Meal Logging
Meal Logging
Input: 📷 Photo of a saladResult: Entry added to Meal Log with:
- Meal: Lunch
- Description: Garden salad
- Estimated calories: ~250
- Photo attached
Recipe Saving
Recipe Saving
Input: https://cooking.nytimes.com/recipes/chocolate-cakeResult: Entry added to Recipes with:
- Title: Chocolate Cake
- Source: NY Times Cooking
- Ingredients: Extracted list
- Instructions: Step-by-step
Workout Logging
Workout Logging
Input: 🎤 “Did 30 minutes on the treadmill, felt good”Result: Entry added to Workouts with:
- Type: Cardio - Treadmill
- Duration: 30 minutes
- Notes: Felt good
Offline Support
Log Anything works offline with some limitations:- Instant Logs ⚡ work fully offline
- Complex entries are queued and sync when online
- Images and voice require connectivity for processing
Tips for Best Results
- Be specific — “Coffee 4” alone is ambiguous
- Include context — “Lunch with clients” helps with categorization
- Use consistent language — Similar inputs route the same way
- Take clear photos — Good lighting helps receipt scanning
- Paste full URLs — Shortened links may not scrape well
- Use scoped input — Tap a Helper chip first for guaranteed 0-credit logging
Credit Usage
| Input Type | Affinity/Local Match | AI-Powered | Fallback |
|---|---|---|---|
| Text | 0 ⚡ | 1 | 0 |
| Voice | — | 9 (8 STT + 1) | 0 |
| Image | — | 5 | 0 |
| Link | — | 1-3 | 0 |
