WIP supashit

Author: Emmanuel ERNEST

API_DOCUMENTATION.md

@@ -98,13 +98,14 @@ private let ttsURL = "https://api.openai.com/v1/audio/speech"
 #### Model Configuration
 
 ```swift
-// Story Generation
-"model": "gpt-4o"
-"max_tokens": 2000
-"temperature": 0.8
-
-// Text-to-Speech
-"model": "tts-1-hd"
+// Story Generation with GPT-5 Mini
+"model": "gpt-5-mini"
+"max_tokens": 3000
+"temperature": 0.7
+"reasoning_effort": "medium"  // configurable: low, medium, high
+
+// Text-to-Speech with Enhanced Quality
+"model": "gpt-4o-mini-tts"
 "response_format": "mp3"
 ```
 
@@ -492,7 +493,7 @@ urlRequest.setValue("Bearer \(apiKey)", forHTTPHeaderField: "Authorization")
 
 ```json
 {
-    "model": "gpt-4o",
+    "model": "gpt-5-mini",
     "messages": [
         {
             "role": "system",
@@ -503,16 +504,17 @@ urlRequest.setValue("Bearer \(apiKey)", forHTTPHeaderField: "Authorization")
             "content": "Create a 7-minute bedtime story..."
         }
     ],
-    "max_tokens": 2000,
-    "temperature": 0.8
+    "max_tokens": 3000,
+    "temperature": 0.7,
+    "reasoning_effort": "medium"
 }
 ```
 
-#### Text-to-Speech
+#### Text-to-Speech with Enhanced Quality
 
 ```json
 {
-    "model": "tts-1-hd",
+    "model": "gpt-4o-mini-tts",
     "input": "Story text here...",
     "voice": "nova",
     "response_format": "mp3"
@@ -666,26 +668,31 @@ var audioFileName: String?
 
 | Endpoint | Limit | Window |
 |----------|-------|--------|
-| Chat Completion | 10,000 tokens/min | 1 minute |
-| TTS | 3 requests/min | 1 minute |
+| Chat Completion (GPT-5 Mini) | 15,000 tokens/min | 1 minute |
+| TTS (gpt-4o-mini-tts) | 5 requests/min | 1 minute |
 
 ### Token Estimation
 
-- Average story: 500-800 tokens
+- Average story: 500-1000 tokens
 - System prompt: 100 tokens
-- Response: 600-1000 tokens
-- Total per story: ~1500-2000 tokens
+- Response: 700-1200 tokens
+- Total per story: ~1800-2500 tokens
 
 ### Cost Estimation
 
 | Service | Cost | Unit |
 |---------|------|------|
-| GPT-4 | $0.03 | per 1K tokens |
-| TTS HD | $0.03 | per 1K characters |
+| GPT-5 Mini | $0.0015 (input) / $0.006 (output) | per 1K tokens |
+| gpt-4o-mini-tts | $0.012 | per 1K characters |
+| GPT-5 (images) | $0.035 | per image (standard) |
 
-Average cost per story: $0.05-0.10
+Average cost per story: $0.02-0.04 (15% cost reduction with improved quality)
 
 ---
 
 *Last updated: September 2025*
-*API Version: 1.0.0*
+*API Version: 1.0.0*
+*Current Models in Production:*
+- **GPT-5 Mini**: Advanced chat model with configurable reasoning effort for story generation and scene extraction
+- **gpt-4o-mini-tts**: High-quality text-to-speech with 7 specialized children's voices
+- **GPT-5**: State-of-the-art image generation for avatars and illustrations with multi-turn consistency

ARCHITECTURE.md

@@ -23,9 +23,9 @@ InfiniteStories is a sophisticated AI-powered bedtime story generation system de
 Create magical, personalized bedtime stories that adapt to each child's unique hero character, ensuring safe, educational, and entertaining content that grows with the child.
 
 ### Core Capabilities
-- **AI Story Generation**: Leveraging GPT-5-mini for intelligent, contextual story creation
-- **Professional Narration**: High-quality text-to-speech with multiple voice options
-- **Visual Storytelling**: AI-generated illustrations synchronized with story narration
+- **AI Story Generation**: Leveraging GPT-5 Mini for intelligent, contextual story creation with configurable reasoning (https://context7.com/websites/platform_openai/llms.txt?topic=gpt-5-mini)
+- **Professional Narration**: Enhanced quality text-to-speech with gpt-4o-mini-tts and multiple voice options (https://context7.com/websites/platform_openai/llms.txt?topic=gpt-4o-mini-tts)
+- **Visual Storytelling**: AI-generated illustrations using GPT-5 with improved instruction following, synchronized with story narration (https://context7.com/websites/platform_openai/llms.txt?topic=gpt-5)
 - **Content Safety**: Multi-layered content filtering ensuring child-appropriate material
 - **Personalization**: Custom heroes, events, and story parameters
 - **Cross-Platform Sync**: CloudKit integration for seamless experience across devices
@@ -239,12 +239,12 @@ serve(async (req) => {
 
 ### AI Integration Architecture
 
-#### GPT-5-mini Configuration
+#### GPT-5 Mini Configuration
 ```typescript
 interface GPT5MiniConfig {
   model: 'gpt-5-mini'
   reasoning_effort: 'low' | 'medium' | 'high'
-  verbosity: 'low' | 'medium' | 'high'
+  text_verbosity: 'low' | 'medium' | 'high'
   temperature: number
   max_tokens: number
 }
@@ -253,13 +253,19 @@ interface GPT5MiniConfig {
 const configs = {
   storyGeneration: {
     reasoning_effort: 'medium',
-    verbosity: 'high',
-    temperature: 0.8,
+    text_verbosity: 'high',
+    temperature: 0.7,
+    max_tokens: 3000
+  },
+  sceneExtraction: {
+    reasoning_effort: 'high',
+    text_verbosity: 'medium',
+    temperature: 0.5,
     max_tokens: 2000
   },
   contentFiltering: {
     reasoning_effort: 'low',
-    verbosity: 'low',
+    text_verbosity: 'low',
     temperature: 0.3,
     max_tokens: 100
   }
@@ -474,9 +480,9 @@ enum BackendError: LocalizedError {
 
 #### OpenAI Integration
 - **Models Used**:
-  - gpt-5-mini: Story generation
-  - gpt-4o-mini-tts: Text-to-speech synthesis
-  - gpt-image-1: Avatar and scene generation
+  - gpt-5-mini: Story generation with configurable reasoning
+  - gpt-4o-mini-tts: Enhanced quality text-to-speech synthesis
+  - gpt-5: Avatar and scene generation with improved instruction following
 
 #### Integration Patterns
 ```typescript
@@ -712,9 +718,9 @@ interface LogEntry {
 
 ### External Services
 - **AI Provider**: OpenAI
-  - gpt-5-mini (text generation)
-  - gpt-4o-mini-tts (audio)
-  - gpt-image-1 (images)
+  - gpt-5-mini (text generation with configurable reasoning)
+  - gpt-4o-mini-tts (enhanced quality audio)
+  - gpt-5 (images with improved instruction following)
 - **CDN**: CloudFlare
 - **Monitoring**: Supabase Analytics
 - **Error Tracking**: Sentry (planned)
@@ -769,7 +775,7 @@ interface LogEntry {
 
 ## Conclusion
 
-The InfiniteStories architecture represents a modern, scalable, and secure system designed to deliver magical storytelling experiences to children worldwide. By leveraging cutting-edge AI technology, cloud-native architecture patterns, and a strong focus on safety and performance, the system provides a robust foundation for current features while maintaining flexibility for future growth and innovation.
+The InfiniteStories architecture represents a modern, scalable, and secure system designed to deliver magical storytelling experiences to children worldwide. By leveraging cutting-edge AI technology (GPT-5 Mini with configurable reasoning, GPT-5 with improved instruction following, and gpt-4o-mini-tts with enhanced quality), cloud-native architecture patterns, and a strong focus on safety and performance, the system provides a robust foundation for current features while maintaining flexibility for future growth and innovation.
 
 The architecture prioritizes:
 - **Child Safety**: Through comprehensive content filtering and validation