Fix new services implementation 5

This commit is contained in:
Urtzi Alfaro
2025-08-15 17:53:59 +02:00
parent 03b4d4185d
commit f7de9115d1
43 changed files with 1714 additions and 891 deletions

View File

@@ -428,7 +428,7 @@ class DataImportService:
repository: SalesRepository,
filename: Optional[str] = None
) -> Dict[str, Any]:
"""Enhanced CSV processing with better data handling"""
"""Enhanced CSV processing with batch product resolution for better reliability"""
try:
reader = csv.DictReader(io.StringIO(csv_content))
rows = list(reader)
@@ -445,22 +445,41 @@ class DataImportService:
# Enhanced column mapping
column_mapping = self._detect_columns(list(rows[0].keys()))
records_created = 0
errors = []
warnings = []
# Pre-process to extract unique products for batch creation
unique_products = set()
parsed_rows = []
logger.info(f"Processing {len(rows)} records from CSV with enhanced mapping")
logger.info(f"Pre-processing {len(rows)} records to identify unique products")
for index, row in enumerate(rows):
try:
# Enhanced data parsing and validation
parsed_data = await self._parse_row_data(row, column_mapping, index + 1)
if parsed_data.get("skip"):
errors.extend(parsed_data.get("errors", []))
warnings.extend(parsed_data.get("warnings", []))
continue
# Resolve product name to inventory_product_id
if not parsed_data.get("skip"):
unique_products.add((
parsed_data["product_name"],
parsed_data.get("product_category", "general")
))
parsed_rows.append((index, parsed_data))
except Exception as e:
logger.warning(f"Failed to parse row {index + 1}: {e}")
continue
logger.info(f"Found {len(unique_products)} unique products, attempting batch resolution")
# Try to resolve/create all unique products in batch
await self._batch_resolve_products(unique_products, tenant_id)
# Now process the actual sales records
records_created = 0
errors = []
warnings = []
logger.info(f"Processing {len(parsed_rows)} validated records for sales creation")
for index, parsed_data in parsed_rows:
try:
# Resolve product name to inventory_product_id (should be cached now)
inventory_product_id = await self._resolve_product_to_inventory_id(
parsed_data["product_name"],
parsed_data.get("product_category"),
@@ -914,47 +933,57 @@ class DataImportService:
logger.info("Import cache cleared for new session")
async def _resolve_product_to_inventory_id(self, product_name: str, product_category: Optional[str], tenant_id: UUID) -> Optional[UUID]:
"""Resolve a product name to an inventory_product_id via the inventory service with caching and rate limiting"""
"""Resolve a product name to an inventory_product_id via the inventory service with improved error handling and fallback"""
# Check cache first
if product_name in self.product_cache:
logger.debug("Product resolved from cache", product_name=product_name, tenant_id=tenant_id)
return self.product_cache[product_name]
# Skip if this product already failed to resolve
# Skip if this product already failed to resolve after all attempts
if product_name in self.failed_products:
logger.debug("Skipping previously failed product", product_name=product_name, tenant_id=tenant_id)
return None
max_retries = 3
base_delay = 1.0 # Start with 1 second delay
max_retries = 5 # Increased retries
base_delay = 2.0 # Increased base delay
fallback_retry_delay = 10.0 # Longer delay for fallback attempts
for attempt in range(max_retries):
try:
# Add delay before API calls to avoid rate limiting
# Add progressive delay to avoid rate limiting
if attempt > 0:
delay = base_delay * (2 ** (attempt - 1)) # Exponential backoff
# Use longer delays for later attempts
if attempt >= 3:
delay = fallback_retry_delay # Use fallback delay for later attempts
else:
delay = base_delay * (2 ** (attempt - 1)) # Exponential backoff
logger.info(f"Retrying product resolution after {delay}s delay",
product_name=product_name, attempt=attempt, tenant_id=tenant_id)
await asyncio.sleep(delay)
# First try to search for existing product by name
products = await self.inventory_client.search_products(product_name, tenant_id)
try:
products = await self.inventory_client.search_products(product_name, tenant_id)
if products:
# Return the first matching product's ID
product_id = products[0].get('id')
if product_id:
uuid_id = UUID(str(product_id))
self.product_cache[product_name] = uuid_id # Cache for future use
logger.info("Resolved product to existing inventory ID",
product_name=product_name, product_id=product_id, tenant_id=tenant_id)
return uuid_id
except Exception as search_error:
logger.warning("Product search failed, trying direct creation",
product_name=product_name, error=str(search_error), tenant_id=tenant_id)
if products:
# Return the first matching product's ID
product_id = products[0].get('id')
if product_id:
uuid_id = UUID(str(product_id))
self.product_cache[product_name] = uuid_id # Cache for future use
logger.info("Resolved product to existing inventory ID",
product_name=product_name, product_id=product_id, tenant_id=tenant_id)
return uuid_id
# Add delay before creation attempt to avoid hitting rate limits
await asyncio.sleep(1.0)
# Add small delay before creation attempt to avoid hitting rate limits
await asyncio.sleep(0.5)
# If not found, create a new ingredient/product in inventory
# If not found or search failed, create a new ingredient/product in inventory
ingredient_data = {
'name': product_name,
'type': 'finished_product', # Assuming sales are of finished products
@@ -965,36 +994,133 @@ class DataImportService:
'category': product_category or 'general'
}
created_product = await self.inventory_client.create_ingredient(ingredient_data, str(tenant_id))
if created_product and created_product.get('id'):
product_id = created_product['id']
uuid_id = UUID(str(product_id))
self.product_cache[product_name] = uuid_id # Cache for future use
logger.info("Created new inventory product for sales data",
product_name=product_name, product_id=product_id, tenant_id=tenant_id)
return uuid_id
try:
created_product = await self.inventory_client.create_ingredient(ingredient_data, str(tenant_id))
if created_product and created_product.get('id'):
product_id = created_product['id']
uuid_id = UUID(str(product_id))
self.product_cache[product_name] = uuid_id # Cache for future use
logger.info("Created new inventory product for sales data",
product_name=product_name, product_id=product_id, tenant_id=tenant_id)
return uuid_id
except Exception as creation_error:
logger.warning("Product creation failed",
product_name=product_name, error=str(creation_error), tenant_id=tenant_id)
logger.warning("Failed to resolve or create product in inventory",
product_name=product_name, tenant_id=tenant_id, attempt=attempt)
except Exception as e:
error_str = str(e)
if "429" in error_str or "rate limit" in error_str.lower():
logger.warning("Rate limit hit, retrying",
if "429" in error_str or "rate limit" in error_str.lower() or "too many requests" in error_str.lower():
logger.warning("Rate limit or service overload detected, retrying with longer delay",
product_name=product_name, attempt=attempt, error=error_str, tenant_id=tenant_id)
if attempt < max_retries - 1:
continue # Retry with exponential backoff
elif "503" in error_str or "502" in error_str or "service unavailable" in error_str.lower():
logger.warning("Service unavailable, retrying with backoff",
product_name=product_name, attempt=attempt, error=error_str, tenant_id=tenant_id)
if attempt < max_retries - 1:
continue # Retry for service unavailable errors
elif "timeout" in error_str.lower() or "connection" in error_str.lower():
logger.warning("Network issue detected, retrying",
product_name=product_name, attempt=attempt, error=error_str, tenant_id=tenant_id)
if attempt < max_retries - 1:
continue # Retry for network issues
else:
logger.error("Error resolving product to inventory ID",
logger.error("Non-retryable error resolving product to inventory ID",
error=error_str, product_name=product_name, tenant_id=tenant_id)
break # Don't retry for non-rate-limit errors
if attempt < max_retries - 1:
# Still retry even for other errors, in case it's transient
continue
else:
break # Don't retry on final attempt
# If all retries failed, mark as failed and return None
# If all retries failed, log detailed error but don't mark as permanently failed yet
# Instead, we'll implement a fallback mechanism
logger.error("Failed to resolve product after all retries, attempting fallback",
product_name=product_name, tenant_id=tenant_id)
# FALLBACK: Try to create a temporary product with minimal data
try:
# Use a simplified approach with minimal data
fallback_data = {
'name': product_name,
'type': 'finished_product',
'unit': 'unit',
'current_stock': 0,
'cost_per_unit': 0
}
logger.info("Attempting fallback product creation with minimal data",
product_name=product_name, tenant_id=tenant_id)
created_product = await self.inventory_client.create_ingredient(fallback_data, str(tenant_id))
if created_product and created_product.get('id'):
product_id = created_product['id']
uuid_id = UUID(str(product_id))
self.product_cache[product_name] = uuid_id
logger.info("SUCCESS: Fallback product creation succeeded",
product_name=product_name, product_id=product_id, tenant_id=tenant_id)
return uuid_id
except Exception as fallback_error:
logger.error("Fallback product creation also failed",
product_name=product_name, error=str(fallback_error), tenant_id=tenant_id)
# Only mark as permanently failed after all attempts including fallback
self.failed_products.add(product_name)
logger.error("Failed to resolve product after all retries",
logger.error("CRITICAL: Permanently failed to resolve product - this will result in missing training data",
product_name=product_name, tenant_id=tenant_id)
return None
async def _batch_resolve_products(self, unique_products: set, tenant_id: str) -> None:
"""Batch resolve/create products to reduce API calls and improve success rate"""
if not unique_products:
return
logger.info(f"Starting batch product resolution for {len(unique_products)} unique products")
# Convert set to list for easier handling
products_list = list(unique_products)
batch_size = 5 # Process in smaller batches to avoid overwhelming the inventory service
for i in range(0, len(products_list), batch_size):
batch = products_list[i:i + batch_size]
logger.info(f"Processing batch {i//batch_size + 1}/{(len(products_list) + batch_size - 1)//batch_size}")
# Process each product in the batch with retry logic
for product_name, product_category in batch:
try:
# Skip if already in cache or failed list
if product_name in self.product_cache or product_name in self.failed_products:
continue
# Try to resolve the product
await self._resolve_product_to_inventory_id(product_name, product_category, tenant_id)
# Add small delay between products to be gentle on the API
await asyncio.sleep(0.5)
except Exception as e:
logger.warning(f"Failed to batch process product {product_name}: {e}")
continue
# Add delay between batches
if i + batch_size < len(products_list):
logger.info("Waiting between batches to avoid rate limiting...")
await asyncio.sleep(2.0)
successful_resolutions = len([p for p, _ in products_list if p in self.product_cache])
failed_resolutions = len([p for p, _ in products_list if p in self.failed_products])
logger.info(f"Batch product resolution completed: {successful_resolutions} successful, {failed_resolutions} failed")
if failed_resolutions > 0:
logger.warning(f"ATTENTION: {failed_resolutions} products failed to resolve - these will be missing from training data")
return
def _structure_messages(self, messages: List[Union[str, Dict]]) -> List[Dict[str, Any]]:
"""Convert string messages to structured format"""
structured = []