payfrit-works/api/setup/analyzeMenuImages.cfm
2026-01-14 22:28:16 -08:00

315 lines
12 KiB
Text

<cfsetting showdebugoutput="false">
<cfsetting enablecfoutputonly="true">
<cfcontent type="application/json; charset=utf-8" reset="true">
<cfscript>
/**
* Analyze Menu Images using Claude Vision API
*
* Accepts uploaded menu images (JPG, PNG, PDF) and sends them to Claude
* to extract structured menu data including:
* - Business info (name, address, phone, hours)
* - Categories
* - Modifier templates
* - Menu items with prices and descriptions
*
* Returns structured JSON for the setup wizard to display
*/
response = { "OK": false };
// Claude API key - should be in environment or config
CLAUDE_API_KEY = application.claudeApiKey ?: "";
// If not in application scope, try to read from config
if (!len(CLAUDE_API_KEY)) {
try {
// Try relative path from current file location first
configPath = getDirectoryFromPath(getCurrentTemplatePath()) & "../../config/claude.json";
if (fileExists(configPath)) {
configData = deserializeJSON(fileRead(configPath));
CLAUDE_API_KEY = configData.apiKey ?: "";
}
// Fallback to expandPath if relative path didn't work
if (!len(CLAUDE_API_KEY)) {
configPath = expandPath("/biz.payfrit.com/config/claude.json");
if (fileExists(configPath)) {
configData = deserializeJSON(fileRead(configPath));
CLAUDE_API_KEY = configData.apiKey ?: "";
}
}
} catch (any e) {
// Config file doesn't exist or is invalid
}
}
if (!len(CLAUDE_API_KEY)) {
response.MESSAGE = "Claude API key not configured";
writeOutput(serializeJSON(response));
abort;
}
try {
// Check for uploaded files
uploadedFiles = [];
// Get form fields for file uploads (file0, file1, file2, etc.)
formFields = form.keyArray();
for (fieldName in formFields) {
if (reFindNoCase("^file[0-9]+$", fieldName) && len(form[fieldName])) {
// This is a file field
arrayAppend(uploadedFiles, fieldName);
}
}
if (arrayLen(uploadedFiles) == 0) {
throw(message="No files uploaded");
}
// Process uploaded files - convert to base64
imageDataArray = [];
// Create temp directory for uploads if needed
uploadDir = getTempDirectory() & "payfrit_menu_uploads/";
if (!directoryExists(uploadDir)) {
directoryCreate(uploadDir);
}
for (fieldName in uploadedFiles) {
try {
// Upload the file
uploadResult = fileUpload(
uploadDir,
fieldName,
"image/jpeg,image/png,image/gif,image/webp,application/pdf",
"makeunique"
);
} catch (any uploadErr) {
throw(message="File upload failed: " & uploadErr.message, detail="Field: " & fieldName);
}
if (uploadResult.fileWasSaved) {
filePath = uploadResult.serverDirectory & "/" & uploadResult.serverFile;
fileExt = lCase(uploadResult.serverFileExt);
// For images, resize if too large (max 1600px on longest side)
if (listFindNoCase("jpg,jpeg,png,gif,webp", fileExt)) {
try {
img = imageRead(filePath);
} catch (any readErr) {
throw(message="Image read failed: " & readErr.message, detail="File: " & filePath);
}
imgWidth = img.width;
imgHeight = img.height;
maxDimension = 1600;
if (imgWidth > maxDimension || imgHeight > maxDimension) {
try {
if (imgWidth > imgHeight) {
newWidth = maxDimension;
newHeight = int(imgHeight * (maxDimension / imgWidth));
} else {
newHeight = maxDimension;
newWidth = int(imgWidth * (maxDimension / imgHeight));
}
imageResize(img, newWidth, newHeight);
} catch (any resizeErr) {
throw(message="Image resize failed: " & resizeErr.message, detail="Dimensions: " & imgWidth & "x" & imgHeight);
}
}
// Re-save with good quality compression
try {
imageWrite(img, filePath, 0.8);
} catch (any writeErr) {
throw(message="Image write failed: " & writeErr.message, detail="File: " & filePath);
}
}
// Read file and convert to base64
fileContent = fileReadBinary(filePath);
base64Content = toBase64(fileContent);
// Determine media type
mediaType = "image/jpeg";
if (fileExt == "png") mediaType = "image/png";
else if (fileExt == "gif") mediaType = "image/gif";
else if (fileExt == "webp") mediaType = "image/webp";
else if (fileExt == "pdf") mediaType = "application/pdf";
// Claude API uses different structure for PDFs vs images
if (fileExt == "pdf") {
arrayAppend(imageDataArray, {
"type": "document",
"source": {
"type": "base64",
"media_type": "application/pdf",
"data": base64Content
}
});
} else {
arrayAppend(imageDataArray, {
"type": "image",
"source": {
"type": "base64",
"media_type": mediaType,
"data": base64Content
}
});
}
// Clean up temp file
fileDelete(filePath);
}
}
if (arrayLen(imageDataArray) == 0) {
throw(message="No valid images could be processed");
}
// Build the prompt for Claude
systemPrompt = "You are an expert at extracting structured menu data from restaurant menu images. Your task is to analyze menu images and return the data in a specific JSON format. Be thorough and extract ALL items, categories, and any modifier patterns you can identify.";
userPrompt = "Please analyze these menu images and extract all the information into a JSON object with these keys: business (with name, address, phone, hours), categories (array with name and itemCount), modifiers (array with name, required boolean, and options array), and items (array with name, description, price, category, and modifiers array). Extract EVERY menu item with accurate prices. Group items by category. Look for modifier patterns like sizes, protein choices, bread choices, and add-ons. Create reusable modifier templates for patterns that appear on multiple items. Return ONLY valid JSON, no other text.";
// Build the messages array with images
messagesContent = [];
// Add each image
for (imgData in imageDataArray) {
arrayAppend(messagesContent, imgData);
}
// Add the text prompt
arrayAppend(messagesContent, {
"type": "text",
"text": userPrompt
});
// Build request body
requestBody = {
"model": "claude-sonnet-4-20250514",
"max_tokens": 8192,
"system": systemPrompt,
"messages": [
{
"role": "user",
"content": messagesContent
}
]
};
// Write request body to temp file (Lucee's HTTP client has issues with large payloads)
tempRequestFile = getTempDirectory() & "claude_request_" & createUUID() & ".json";
tempResponseFile = getTempDirectory() & "claude_response_" & createUUID() & ".json";
fileWrite(tempRequestFile, serializeJSON(requestBody));
try {
// Call Claude API using cfhttp tag (more reliable than http component)
httpResult = {};
http url="https://api.anthropic.com/v1/messages" method="POST" timeout=300 result="httpResult" {
httpparam type="header" name="Content-Type" value="application/json";
httpparam type="header" name="x-api-key" value=CLAUDE_API_KEY;
httpparam type="header" name="anthropic-version" value="2023-06-01";
httpparam type="body" value=serializeJSON(requestBody);
}
httpStatusCode = httpResult.statusCode ?: "0";
httpFileContent = httpResult.fileContent ?: "";
// Normalize status code
if (isNumeric(httpStatusCode)) {
httpStatusCode = int(httpStatusCode);
} else if (findNoCase("200", httpStatusCode)) {
httpStatusCode = 200;
}
if (httpStatusCode != 200) {
// Try to parse error message from Claude's response
errorMsg = "Claude API error: " & httpResult.statusCode;
try {
errorData = deserializeJSON(httpFileContent);
if (structKeyExists(errorData, "error") && structKeyExists(errorData.error, "message")) {
errorMsg = errorMsg & " - " & errorData.error.message;
}
} catch (any e) {
// Use raw response if can't parse
}
throw(message=errorMsg, detail=httpFileContent);
}
} finally {
// Clean up temp files
if (fileExists(tempRequestFile)) fileDelete(tempRequestFile);
}
// Parse Claude's response
claudeResponse = deserializeJSON(httpFileContent);
if (!structKeyExists(claudeResponse, "content") || !arrayLen(claudeResponse.content)) {
throw(message="Empty response from Claude");
}
// Extract the text content
responseText = "";
for (block in claudeResponse.content) {
if (block.type == "text") {
responseText = block.text;
break;
}
}
// Parse the JSON from Claude's response
// Claude might wrap it in markdown code blocks, so strip those
responseText = trim(responseText);
if (left(responseText, 7) == "```json") {
responseText = mid(responseText, 8, len(responseText) - 7);
}
if (left(responseText, 3) == "```") {
responseText = mid(responseText, 4, len(responseText) - 3);
}
if (right(responseText, 3) == "```") {
responseText = left(responseText, len(responseText) - 3);
}
responseText = trim(responseText);
extractedData = deserializeJSON(responseText);
// Update category item counts
if (structKeyExists(extractedData, "categories") && structKeyExists(extractedData, "items")) {
for (i = 1; i <= arrayLen(extractedData.categories); i++) {
catName = extractedData.categories[i].name;
itemCount = 0;
for (item in extractedData.items) {
if (structKeyExists(item, "category") && item.category == catName) {
itemCount++;
}
}
extractedData.categories[i].itemCount = itemCount;
}
}
// Add unique IDs to items
if (structKeyExists(extractedData, "items")) {
for (i = 1; i <= arrayLen(extractedData.items); i++) {
extractedData.items[i].id = "item_" & i;
}
}
response.OK = true;
response.DATA = extractedData;
response.imagesProcessed = arrayLen(imageDataArray);
} catch (any e) {
response.MESSAGE = e.message;
if (len(e.detail)) {
response.DETAIL = e.detail;
}
// Add debug info
if (structKeyExists(e, "tagContext") && arrayLen(e.tagContext) > 0) {
response.DEBUG_LINE = e.tagContext[1].line;
response.DEBUG_TEMPLATE = e.tagContext[1].template;
}
}
writeOutput(serializeJSON(response));
</cfscript>