Upload app.py

app.py

@@ -1002,27 +1002,13 @@ with gr.Blocks(title="OpenEvolve Prompt Optimizer", theme=gr.themes.Soft()) as d
     gr.Markdown("""
     # 🧬 OpenEvolve Prompt Optimizer
 
-    Automatically evolve and optimize your prompts using evolutionary algorithms!
-
-    This space uses [OpenEvolve](https://github.com/algorithmicsuperintelligence/openevolve) to iteratively improve prompts
-    by testing them on real datasets and evolving better versions.
-
-    ## 🔑 Setup (Required)
-    **To use this space:**
-    1. Click "⋮" (menu) → "Duplicate Space" to create your own copy
-    2. In your duplicated space, go to Settings → Variables & Secrets
-    3. Add your OpenRouter API key as `OPENAI_API_KEY`
-    4. Get a free API key at [openrouter.ai](https://openrouter.ai/)
-
-    ## How it works:
-    1. Enter an initial prompt (use `{input}` as a placeholder for dataset inputs)
-    2. Default dataset is **IMDB** (movie review sentiment classification) - great for showing prompt improvement!
-    3. Specify the dataset split and field names (or use other datasets like `gsm8k`, `dair-ai/emotion`)
-    4. Click "Optimize Prompt" - the system will validate everything first!
-    5. Watch the evolution progress in real-time
-    6. Compare initial vs. best prompt side-by-side (both evaluated on same 50 samples)!
-
-    **Model**: Using `meta-llama/llama-3.2-3b-instruct` (reliable, very cheap at ~$0.04 per 1M tokens)
+    Automatically optimize prompts using evolutionary algorithms. Evolves better prompts by testing on real datasets.
+
+    **Setup**: Duplicate this space, add your OpenRouter API key (`OPENAI_API_KEY`) in Settings → Secrets. Get free key at [openrouter.ai](https://openrouter.ai/)
+
+    **Usage**: Enter initial prompt with `{input}` placeholder → Click optimize → Compare results
+
+    **Model**: `meta-llama/llama-3.2-3b-instruct` (~$0.04 per 1M tokens)
     """)
 
     with gr.Row():
@@ -1086,49 +1072,6 @@ with gr.Blocks(title="OpenEvolve Prompt Optimizer", theme=gr.themes.Soft()) as d
         with gr.Column():
             final_results = gr.Markdown("### Best Prompt\nWill appear here after optimization...", visible=True)
 
-    # Documentation section - in collapsible accordion
-    gr.Markdown("---")
-    with gr.Accordion("📚 Documentation & Examples", open=False):
-        gr.Markdown("""
-        ### Example Datasets & Fields:
-
-        | Dataset | Split | Input Field | Target Field | Task | Size |
-        |---------|-------|-------------|--------------|------|------|
-        | **stanfordnlp/imdb** | test | text | label | Sentiment Analysis | 25,000 |
-        | dair-ai/emotion | test | text | label | Emotion Classification | 2,000 |
-        | gsm8k | train | question | answer | Grade School Math | 7,473 |
-        | fancyzhx/ag_news | test | text | label | News Classification | 7,600 |
-        | MathArena/aime_2025 | train | problem | answer | Hard Math (AIME) | 30 |
-
-        ### About This Demo Space:
-
-        **This is a demonstration space** showcasing OpenEvolve's prompt optimization capabilities.
-        The interface shows you how the system works, but **you'll need to set up your own instance to run optimizations**.
-
-        ### How to Run This Yourself:
-
-        1. **Clone this Space**: Click "⋮" (three dots) at top-right → "Duplicate this Space"
-        2. **Set Environment Variables** in your cloned Space's settings:
-           - `OPENAI_API_KEY`: Your OpenRouter API key (get free key at [openrouter.ai/keys](https://openrouter.ai/keys))
-           - `HF_TOKEN`: (Optional) HuggingFace token for private datasets
-        3. **Configure Your Optimization**:
-           - Dataset: Use full name format (e.g., `stanfordnlp/imdb` or `openai/gsm8k`)
-           - Fields: Specify exact field names from the dataset schema
-           - Model: Choose from 5 curated free models (larger models = better results but slower/rate-limited)
-        4. **Run & Monitor**:
-           - All inputs are validated before starting
-           - **Evaluation strategy**:
-             - Initial evaluation: 200 samples
-             - Final evaluation: Same 200 samples (fair comparison)
-             - Evolution: Each variant tested on same 200 samples (ensures optimization aligns with test set)
-           - Compare initial vs best prompt side-by-side with identical test sets
-
-        ### About OpenEvolve:
-        OpenEvolve is an open-source evolutionary optimization framework. Learn more at:
-        - [GitHub Repository](https://github.com/algorithmicsuperintelligence/openevolve)
-        - [Documentation](https://github.com/algorithmicsuperintelligence/openevolve#readme)
-        """)
-
     # Wire up the optimize button with hardcoded model
     def optimize_with_fixed_model(initial_prompt, dataset_name, dataset_split,
                                    input_field, target_field, progress=gr.Progress()):