ARCHITECTURE FIX:
- aws-admin IntegrationSettings will NEVER exist (it's a legacy pattern)
- Only user's own account IntegrationSettings can exist (if they override defaults)
- Otherwise GlobalIntegrationSettings is used directly
- API keys are ALWAYS from GlobalIntegrationSettings (accounts cannot override API keys)
REMOVED:
- All aws-admin Account lookups
- All aws-admin IntegrationSettings fallback attempts
- Confusing nested try/except chains
CORRECT FLOW NOW:
1. Try account's IntegrationSettings for config overrides
2. Use GlobalIntegrationSettings for missing values and ALL API keys
3. No intermediate aws-admin lookups
ROOT CAUSES IDENTIFIED:
1. GlobalAIPrompt template had unescaped JSON braces that broke Python's .format()
- Python treats {...} as placeholders, causing KeyError when rendering
- Escaped JSON braces to {{...}} while preserving {title}, {content}, {max_images}
2. Image functions hardcoded aws-admin IntegrationSettings which didn't exist
- Functions failed when aws-admin account had no IntegrationSettings
- Added GlobalIntegrationSettings fallback for all missing values
CHANGES:
- Fixed GlobalAIPrompt.image_prompt_extraction template in database (escaped JSON)
- Updated generate_image_prompts._get_max_in_article_images() with fallback
- Updated generate_images.prepare_data() with fallback for all image settings
- Updated tasks.process_image_generation_queue() with fallback for config + API keys
TESTED: Template rendering now works, GlobalIntegrationSettings.max_in_article_images=4
- Added AIModelConfig model to manage AI model configurations in the database.
- Created serializers and views for AI model configurations, enabling read-only access via REST API.
- Implemented filtering capabilities for model type, provider, and default status in the API.
- Seeded initial data for text and image models, including pricing and capabilities.
- Updated Django Admin interface for managing AI models with enhanced features and bulk actions.
- Added validation methods for model and image size checks.
- Comprehensive migration created to establish the AIModelConfig model and seed initial data.
- Documented implementation and validation results in summary and report files.
- Backfills model_config FK from model_name in CreditUsageLog
- Calculates cost_usd_input/output/total from tokens and model pricing
- Processes in configurable batches (default 500)
- Includes dry-run mode for safety
- Successfully backfilled 250 historical logs
- Enables full token analytics on historical data
- Reverted to commit #10 (98e68f6) for stable AI function base
- Fixed database migrations: removed 0018-0019 that broke schema
- Fixed CreditCostConfig schema: restored credits_cost, unit fields
- Fixed historical table schema for django-simple-history
- Added debug system (staged for future use)
Changes:
- CreditCostConfig: Updated OPERATION_TYPE_CHOICES (10 ops, no duplicates)
- CreditUsageLog: Updated choices with legacy aliases marked
- Migration 0018_update_operation_choices: Applied successfully
- All AI operations working (clustering, ideas, content, optimization, etc.)
Test Results:
✓ CreditCostConfig save/load working
✓ Credit check passing for all operations
✓ AICore initialization successful
✓ AIEngine operation mapping functional
✓ Admin panel accessible without 500 errors
Future: AI-MODEL-COST-REFACTOR-PLAN.md created for token-based system
- Introduced GlobalModuleSettings model for platform-wide module enable/disable settings.
- Added 'caption' field to Images model to store image captions.
- Updated GenerateImagePromptsFunction to handle new caption structure in prompts.
- Enhanced AIPromptViewSet to return global prompt types and validate active prompts.
- Modified serializers and views to accommodate new caption field and global settings.
- Updated frontend components to display captions and filter prompts based on active types.
- Created migrations for GlobalModuleSettings and added caption field to Images.
feat(admin): Add API monitoring, debug console, and system health templates for enhanced admin interface
docs: Add AI system cleanup summary and audit report detailing architecture, token management, and recommendations
docs: Introduce credits and tokens system guide outlining configuration, data flow, and monitoring strategies
- Updated AutomationService to include estimated_word_count.
- Increased stage_1_batch_size from 20 to 50 in AutomationViewSet.
- Changed Keywords model to replace 'intent' property with 'country'.
- Adjusted ClusteringService to allow a maximum of 50 keywords for clustering.
- Modified admin and management commands to remove 'intent' and use 'country' instead.
- Updated serializers to reflect the change from 'intent' to 'country'.
- Adjusted views and filters to use 'country' instead of 'intent'.
- Updated frontend forms, filters, and pages to replace 'intent' with 'country'.
- Added migration to remove 'intent' field and add 'country' field to SeedKeyword model.
