Switch backend from Python/Flask to Rust/Axum

https://claude.ai/code/session_015myTTMs6yDsAGarATe5ePZ

PLAN.md

@@ -14,9 +14,9 @@
 
 | Layer       | Technology                  |
 |-------------|-----------------------------|
-| Backend     | Python + Flask              |
+| Backend     | Rust + Axum                 |
 | AI / OCR    | Claude API (vision)         |
-| Spreadsheet | openpyxl (Excel .xlsx)      |
+| Spreadsheet | rust_xlsxwriter (Excel)     |
 | Frontend    | Mobile-friendly HTML        |
 
 ---
@@ -33,11 +33,13 @@
 
 ```
 Driverthing/
-├── app.py              # Flask server + Claude API call
-├── mileage.py          # Excel read/write logic
+├── src/
+│   ├── main.rs         # Axum server + routes
+│   ├── claude.rs       # Claude API vision call
+│   └── excel.rs        # Excel read/write logic
 ├── templates/
 │   └── index.html      # Mobile camera upload page
-├── requirements.txt
+├── Cargo.toml
 └── mileage_log.xlsx    # Generated, gitignored
 ```
 
@@ -47,7 +49,7 @@ Driverthing/
 
 ```
 Phone camera → upload photo
-    → Flask receives image
+    → Axum receives multipart/form-data
         → Claude Vision: "What is the odometer reading?"
             → Extract number
                 → Calculate delta from last reading
@@ -59,11 +61,11 @@ Phone camera → upload photo
 
 ## Implementation steps
 
-1. Set up Flask app with a file upload endpoint
-2. Send uploaded image to Claude API with a vision prompt
+1. Set up Axum server with a multipart file upload endpoint
+2. Send uploaded image (base64) to Claude API with a vision prompt
 3. Parse the odometer number from the response
 4. Read the last recorded odometer value from the Excel file
 5. Calculate the difference (km driven)
 6. Append a new row (date, time, odometer, trip km, notes) to the sheet
 7. Return confirmation to the browser
-8. Add a download button for the Excel file
+8. Add a download endpoint for the Excel file