Consolidate all fixes into v0.9.3 and synchronize version strings

Author: x-hannibal

CHANGELOG.md

@@ -1,42 +1,42 @@
-* 2026-02-16: v0.9.4 - Aspect Ratio & CLI Robustness (Hannibal)  
-  * Fixed a critical bug where the default Aspect Ratio User Valve was overriding custom Size settings.
-  * Implemented typographic dash normalization in the Prompt Parser to support mobile/iOS auto-corrections.
-  * Updated documentation with detailed configuration hierarchy and parameter behavior.
+* 2026-02-16: v0.9.3 - Aspect Ratio, CLI Robustness & Privacy Documentation (Hannibal)  
 
-* 2026-02-16: v0.9.3 - Documentation & UX Clarity (Hannibal)  
-  * Updated README.md to clarify the configuration hierarchy between CLI, User Valves, Admin Valves, and Global OWUI Settings.
-  * Refined explanation of the "Size" parameter location to prevent user confusion.
+  * Fixed a critical bug where the default Aspect Ratio User Valve was overriding custom Size settings.
+  * Implemented typographic dash normalization in the Prompt Parser to support mobile/iOS auto-corrections (converting — and – back to standard --).
+  * Added a dedicated "Privacy & Local Execution" section to the README explaining how to use `easy_cloud_mode` for local-only setups (LM Studio/Local Forge).
+  * Updated documentation with a detailed configuration hierarchy and parameter behavior.
+  * Refined the "Size" parameter location explanation to guide users to the Chat Controls menu.
 
 * 2026-02-15: v0.9.2-beta.2 - Multilingual Prompt Enhancer & Vision QC (Hannibal)  
+
   * Subcommand Architecture: Unified workflow with `img:p` (Prompt Only), `img:r` (Entropy Randomizer), and `img ?` (Interactive Help).
-  * Entropy Engine 2.0: Replaced LLM-based randomizer with a deterministic Python-side "Double Dice" system (20 Categories + 10 Moods) to enforce physical grounding and stylistic diversity.
-  * Direct Dispatch Core: Implemented a global persistent HTTP client for zero-latency connection pooling and direct backend execution (bypassing OWUI chat overhead).
-  * Smart VRAM Management: Added automatic model unloading and an "Extreme Cleanup" valve to prevent OOM errors on local backends.
+  * Entropy Engine 2.0: Replaced LLM-based randomizer with a deterministic Python-side "Double Dice" system (20 Categories + 10 Moods).
+  * Direct Dispatch Core: Implemented a global persistent HTTP client for zero-latency connection pooling.
+  * Smart VRAM Management: Added automatic model unloading and an "Extreme Cleanup" valve.
   * Forge Power-Ups: Exposed granular High-Res Fix controls via CLI (`hr`, `hru`, `dns`, `hdcs`).
-  * Error Intelligence: Now extracts and displays raw API error messages (e.g., OpenAI Safety Violations) directly in the chat.
-  * Fixes: Resolved OpenAI size snapping (Error 400), Global Config retrieval for `PersistentConfig` objects, and unified Forge/A1111 terminology.
+  * Error Intelligence: Extracts and displays raw API error messages (e.g., OpenAI Safety Violations).
+  * Technical Fixes: Resolved OpenAI size snapping, Global Config retrieval for PersistentConfig, and Forge terminology.
 
 * 2026-02-10: v0.8.1 - Intelligent Router Refactor (Hannibal)  
+
   * Intelligent Router: Implemented a dispatcher with dedicated methods for each engine.
   * Native OpenAI & Gemini: Added direct HTTPX support for OpenAI and Gemini with full parameter mapping.
-  * Advanced Mapping: Implemented logic for `style`/`quality` (OpenAI) and `seed`/`watermark` (Gemini).
   * UI Cleanup: Improved code block formatting in chat messages.
 
 * 2026-02-08: v0.7.5 - SoC Refactor & Inference Optimization (Hannibal)  
+
   * SoC Refactor: Restructured the code into specialized classes (`InferenceEngine`, `PromptParser`, etc.).
   * Optimized Inference: Generalized the `_infer` method to handle text and vision calls agnostically.
-  * Local Performance Boost: Optimized the flow to skip the vision probe when using `imgx`.
 
 * 2026-01-27: v0.6.3 - Modular Architecture & Vision Audit (Hannibal)  
+
   * Modular Architecture: Complete refactor to a method-based class structure for better maintainability.
-  * Precise Metrics: Replaced heuristic token estimation with exact data from Open WebUI API responses.
   * Vision Cache Persistence: Introduced `persistent_vision_cache` to store model capabilities locally.
-  * Audit Optimization: Refined Vision Audit prompts for objective scoring and artifact detection.
 
 * 2026-01-25: v0.5.1 - Multimodal Validation & Debugging (Hannibal)  
+
   * Multimodal Validation: Implemented base64 reference image probing to verify vision capabilities.
   * Debug Mode: Added a toggle for detailed console logging.
 
 * 2026-01-23: v0.4.83 - Initial Vision Audit Integration (Hannibal)  
-  * Vision Audit System: Initial integration of real-time image analysis.
-  * Unicode UI: Optimized technical details for environments with restricted Markdown.
+
+  * Vision Audit System: Initial integration of real-time image analysis.

README.md

@@ -10,11 +10,11 @@ This filter acts as an **Intelligent Dispatcher**, unlocking advanced, engine-sp
 ![License](https://img.shields.io/github/license/annibale-x/Easymage?color=green)
 
 
-### 🆕 What's New in v0.9.4 (vs v0.9.3)
+### 🆕 What's New in v0.9.3 (vs v0.9.2-beta.2)
 
--   **Aspect Ratio Intelligence**: Fixed a critical bug where a hidden default aspect ratio was overriding custom dimensions. The system now mathematically infers the correct ratio from your `Size` setting automatically. It prioritizes explicit `Aspect Ratio` values only when manually set, recalculating the image height to ensure perfect proportions.
--   **CLI Typographic Normalization**: The Prompt Parser now intercepts and normalizes typographic dashes (like the em-dash `—` or en-dash `–` automatically inserted by mobile keyboards and macOS/iOS) back into valid CLI parameters (`--`), preventing syntax errors on smartphones.
--   **Privacy & Local Setup Documentation**: Added a dedicated section explaining how to force 100% local execution using the `easy_cloud_mode` toggle, specifically for users running backends like LM Studio or local ComfyUI.
+-   **Aspect Ratio Intelligence**: Fixed a critical bug where a hidden default aspect ratio was overriding custom dimensions. The system now mathematically infers the correct ratio from your `Size` setting automatically.
+-   **CLI Typographic Normalization**: The Prompt Parser now intercepts and normalizes typographic dashes (like the em-dash `—` or en-dash `–` automatically inserted by mobile keyboards) back into valid CLI parameters (`--`), preventing syntax errors on smartphones.
+-   **Privacy & Local Setup**: Added a dedicated section explaining how to force 100% local execution using the `easy_cloud_mode` toggle.
 
 ---
 

__pycache__/easymage.cpython-310.pyc

@@ -1,6 +1,6 @@
 """
 title: ✨Easymage: Generative Imaging & Prompt Engineering Filter
-version: 0.9.4
+version: 0.9.3
 repo_url: https://github.com/annibale-x/Easymage
 author: Hannibal
 author_url: https://openwebui.com/u/h4nn1b4l
@@ -24,7 +24,7 @@
 from pydantic import BaseModel, Field
 
 EM_ICON = "✨"
-EM_VERSION = "0.9.4"
+EM_VERSION = "0.9.3"
 CAPABILITY_CACHE_PATH = "data/easymage_vision_cache.json"
 
 # --- GLOBAL SERVICES ---