llm feedack updates (#94)

* Simplify LLM feedback code while maintaining core functionality

- Streamlined UI with cleaner layout and better visual hierarchy
- Reduced CSS by 70% using a simpler container-based approach
- Simplified JavaScript from ~400 lines to ~130 lines
- Improved state management and evolution logic
- Maintained all core functionality including art generation, preview, and history

Closes #91

Mentat precommits passed. Log: https://mentat.ai/log/ccb24099-4857-47b7-b5d7-d856a21c9041

* Update .gitignore with additional common patterns

Added patterns for:
- Yarn PnP files
- Additional build output directory (out/)
- General debug logs
- OS-specific files (DS_Store, Thumbs.db)

Mentat precommits passed. Log: https://mentat.ai/log/75248827-ef9b-4de6-9167-557d758749fb

* Make .gitignore more comprehensive for all app types

- Removed wildcard prefixes to work recursively at any depth
- Added patterns for various build tools and frameworks
- Added web-specific patterns for compiled assets
- Added testing and temporary file patterns
- Improved organization and categorization

Mentat precommits passed. Log: https://mentat.ai/log/9478638d-b20e-4b73-a8ec-7dc7c7cfd202

* Update .gitignore to better handle subfolder apps

- Added `**/` prefix to match files at any directory depth
- Added patterns for Yarn 2+ cache files
- Included vanilla HTML/JS specific patterns
- Added more build output and test directories
- Improved handling of OS-specific files at all levels

Mentat precommits passed. Log: https://mentat.ai/log/4786f47d-56c8-44b6-bb53-1e2e7696062a

* Revert "Make .gitignore more comprehensive for all app types"

This reverts commit a424654.

* enhance visual poetry UI and generation

enhance visual poetry UI and generation

- Update presets to focus on visual poetry styles (Apollinaire, Mallarmé, etc)
- Improve monospace text display with better styling and dimensions
- Add seed parameter for reproducible generations
- Remove separate current art window in favor of preview
- Enhance generation logic with better state management
- Improve error handling and logging

* Tweaks more prompts

* add model selector

* more robust fetching logic

---------

Co-authored-by: MentatBot <160964065+MentatBot@users.noreply.github.com>

Author: voodoohop

llm-feedback/index.html

@@ -32,7 +32,7 @@
             font-family: 'Courier New', Courier, monospace;
             white-space: pre;
             display: block;
-            width: 480px;
+            width: 720px;
             height: 600px;
             overflow: auto;
             background: #f8f8f8;
@@ -252,11 +252,25 @@ <h2>History</h2>
                 let seed = elements.seed.value ? parseInt(elements.seed.value) : Math.floor(Math.random() * 10);
                 elements.seed.value = seed; // Update input with chosen seed
 
-                const url = `https://text.pollinations.ai/${encodedPrompt}?system=${encodedSystem}&seed=${seed}&model=${getSelectedModel()}`;
-                console.log('Request URL:', url);
+
 
+                let response = null;
                 console.log('Fetching...');
-                const response = await fetch(url);
+                do {
+                    const url = `https://text.pollinations.ai/${encodedPrompt}?system=${encodedSystem}&seed=${seed}&model=${getSelectedModel()}`;
+                    console.log('Request URL:', url);
+
+                    try {
+                        response = await fetch(url);
+                    } catch (error) {
+                        console.error('Fetch error:', error);
+                    }
+                    if (!response.ok) {
+                        seed = seed + 1;
+                        await new Promise(resolve => setTimeout(resolve, 10000));
+                    }
+                    console.log('Trying fetch...', url);
+                } while (!response.ok);
                 console.log('Response received:', response.status);
 
                 const text = await response.text();