Add Kimi model support and enhance LLM comparison tool functionality

- Introduced new Kimi model configurations with pricing and supported parameters in the MODELS_CONFIG.
- Implemented the call_kimi function to handle API requests for Kimi models, including cost calculation based on token usage.
- Updated the LLMClient class to route calls to Kimi models.
- Enhanced the UI to include Kimi API key input and updated model addition logic to accommodate the new provider.
- Improved results formatting and added functionality for toggling fullscreen mode and filtering results to hide errors.

app.py

@@ -68,6 +68,19 @@ MODELS_CONFIG = {
         "qwen3-235b-a22b-thinking-2507": {"input_price": 0.11, "output_price": 0.6, "supports": ["temperature", "top_p", "max_tokens", "top_k"]},
         "qwq-32b": {"input_price": 0.15, "output_price": 0.4, "supports": ["temperature", "top_p", "max_tokens", "top_k"]},
     },
+    "Kimi": {
+        "kimi-k2-0905-preview": {"input_price": 0.50, "output_price": 2.00, "supports": ["temperature", "top_p", "max_tokens"]},
+        "kimi-k2-0711-preview": {"input_price": 0.50, "output_price": 2.00, "supports": ["temperature", "top_p", "max_tokens"]},
+        "kimi-k2-turbo-preview": {"input_price": 0.30, "output_price": 1.20, "supports": ["temperature", "top_p", "max_tokens"]},
+        "moonshot-v1-8k": {"input_price": 0.12, "output_price": 0.12, "supports": ["temperature", "top_p", "max_tokens"]},
+        "moonshot-v1-32k": {"input_price": 0.24, "output_price": 0.24, "supports": ["temperature", "top_p", "max_tokens"]},
+        "moonshot-v1-128k": {"input_price": 0.60, "output_price": 0.60, "supports": ["temperature", "top_p", "max_tokens"]},
+        "moonshot-v1-auto": {"input_price": 0.24, "output_price": 0.24, "supports": ["temperature", "top_p", "max_tokens"]},
+        "kimi-latest": {"input_price": 0.50, "output_price": 2.00, "supports": ["temperature", "top_p", "max_tokens"]},
+        "moonshot-v1-8k-vision-preview": {"input_price": 0.15, "output_price": 0.15, "supports": ["temperature", "top_p", "max_tokens"]},
+        "moonshot-v1-32k-vision-preview": {"input_price": 0.30, "output_price": 0.30, "supports": ["temperature", "top_p", "max_tokens"]},
+        "moonshot-v1-128k-vision-preview": {"input_price": 0.70, "output_price": 0.70, "supports": ["temperature", "top_p", "max_tokens"]},
+    },
 }
 
 @dataclass
@@ -357,6 +370,43 @@ class LLMClient:
         except Exception as e:
             return None, 0, 0, str(e)
     
+    async def call_kimi(self, model: str, prompt: str, config: ModelConfig) -> tuple:
+        """Call Kimi/Moonshot API (OpenAI-compatible)"""
+        try:
+            import openai
+            client = openai.AsyncOpenAI(
+                api_key=self.api_keys.get("Kimi", ""),
+                base_url="https://api.moonshot.cn/v1"
+            )
+            
+            start_time = time.time()
+            
+            params = {
+                "model": model,
+                "messages": [{"role": "user", "content": prompt}],
+                "temperature": config.temperature,
+                "top_p": config.top_p,
+                "max_tokens": config.max_tokens,
+            }
+            
+            response = await client.chat.completions.create(**params)
+            
+            elapsed_time = time.time() - start_time
+            output = response.choices[0].message.content
+            
+            # Calculate tokens and cost
+            input_tokens = response.usage.prompt_tokens if hasattr(response, 'usage') else len(prompt.split()) * 1.3
+            output_tokens = response.usage.completion_tokens if hasattr(response, 'usage') else len(output.split()) * 1.3
+            
+            model_info = MODELS_CONFIG["Kimi"][model]
+            cost = (input_tokens * model_info["input_price"] + output_tokens * model_info["output_price"]) / 1_000_000
+            cost_per_1000 = cost * 1000
+            
+            return output, elapsed_time, cost_per_1000, None
+            
+        except Exception as e:
+            return None, 0, 0, str(e)
+    
     async def call_model(self, provider: str, model: str, prompt: str, config: ModelConfig) -> tuple:
         """Route to appropriate API call"""
         # Check if API key is provided
@@ -378,6 +428,8 @@ class LLMClient:
             return await self.call_deepseek(model, prompt, config)
         elif provider == "Qwen":
             return await self.call_qwen(model, prompt, config)
+        elif provider == "Kimi":
+            return await self.call_kimi(model, prompt, config)
         else:
             return None, 0, 0, f"Unknown provider: {provider}"
 
@@ -385,7 +437,7 @@ class LLMClient:
 selected_models = []
 
 def add_model(provider, model, temperature, top_p, max_tokens, top_k, freq_penalty, pres_penalty, in_price, out_price, 
-              openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key):
+              openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, kimi_key):
     """Add a model to the comparison list"""
     if not provider or not model:
         return create_model_list_display(), "⚠️ Please select both provider and model"
@@ -399,6 +451,7 @@ def add_model(provider, model, temperature, top_p, max_tokens, top_k, freq_penal
         "Mistral": mistral_key,
         "Deepseek": deepseek_key,
         "Qwen": qwen_key,
+        "Kimi": kimi_key,
     }
     
     added_count = 0
@@ -592,6 +645,7 @@ def update_on_provider_change(provider):
             gr.Textbox(visible=True, info="Get key: https://console.mistral.ai/"),
             gr.Textbox(visible=True, info="Get key: https://platform.deepseek.com/"),
             gr.Textbox(visible=True, info="Get key: https://dashscope.console.aliyun.com/"),
+            gr.Textbox(visible=True, info="Get key: https://platform.moonshot.cn/"),
         ]
         in_price = 0.0
         out_price = 0.0
@@ -610,6 +664,7 @@ def update_on_provider_change(provider):
             gr.Textbox(visible=provider == "Mistral", info="Get key: https://console.mistral.ai/"),
             gr.Textbox(visible=provider == "Deepseek", info="Get key: https://platform.deepseek.com/"),
             gr.Textbox(visible=provider == "Qwen", info="Get key: https://dashscope.console.aliyun.com/"),
+            gr.Textbox(visible=provider == "Kimi", info="Get key: https://platform.moonshot.cn/"),
         ]
         
         # Get pricing from first model as default
@@ -619,7 +674,7 @@ def update_on_provider_change(provider):
         out_price = model_info["output_price"]
     else:
         model_dropdown_update = gr.Dropdown(choices=[], value=None)
-        api_keys = [gr.Textbox(visible=False)] * 7
+        api_keys = [gr.Textbox(visible=False)] * 8
         in_price = 0.0
         out_price = 0.0
     
@@ -670,7 +725,7 @@ def update_parameter_visibility(provider, model):
         gr.Number(visible="presence_penalty" in supported),
     ]
 
-async def run_comparison(prompt, openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key):
+async def run_comparison(prompt, openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, kimi_key):
     """Run comparison across all selected models"""
     if not selected_models:
         return None, "⚠️ Please add at least one model first", gr.update(visible=False)
@@ -686,6 +741,7 @@ async def run_comparison(prompt, openai_key, anthropic_key, google_key, cohere_k
         "Mistral": mistral_key,
         "Deepseek": deepseek_key,
         "Qwen": qwen_key,
+        "Kimi": kimi_key,
     }
     
     client = LLMClient(api_keys)
@@ -720,21 +776,30 @@ async def run_comparison(prompt, openai_key, anthropic_key, google_key, cohere_k
             cost = (input_tokens * input_price + output_tokens * output_price) / 1_000_000
             cost_per_1000 = cost * 1000
         
-        # Format model parameters
-        params_str = f"temp={model_config['temperature']}, top_p={model_config['top_p']}, max_tokens={model_config['max_tokens']}"
+        # Format model name in 3 lines: Provider-Model, Parameters, Costs
+        # Line 1: Provider - Model
+        line1 = f"**{model_config['provider']} - {model_config['model']}**"
+        
+        # Line 2: Parameters
+        params_list = [
+            f"temp={model_config['temperature']}", 
+            f"top_p={model_config['top_p']}", 
+            f"max_tokens={model_config['max_tokens']}"
+        ]
         if model_config.get('top_k'):
-            params_str += f", top_k={model_config['top_k']}"
+            params_list.append(f"top_k={model_config['top_k']}")
         if model_config.get('frequency_penalty'):
-            params_str += f", freq_penalty={model_config['frequency_penalty']}"
+            params_list.append(f"freq_pen={model_config['frequency_penalty']}")
         if model_config.get('presence_penalty'):
-            params_str += f", pres_penalty={model_config['presence_penalty']}"
+            params_list.append(f"pres_pen={model_config['presence_penalty']}")
+        line2 = ", ".join(params_list)
         
-        # Add pricing info to model name
+        # Line 3: Costs
         input_price = model_config.get('input_price', 0)
         output_price = model_config.get('output_price', 0)
-        params_str += f"\n💰 ${input_price:.2f}/${output_price:.2f} per 1M tokens"
+        line3 = f"💰 ${input_price:.2f}/${output_price:.2f} per 1M tokens"
         
-        model_name = f"{model_config['provider']} - {model_config['model']}\n({params_str})"
+        model_name = f"{line1}  \n{line2}  \n{line3}"
         
         if error:
             results.append({
@@ -744,13 +809,22 @@ async def run_comparison(prompt, openai_key, anthropic_key, google_key, cohere_k
                 "Output": f"❌ Error: {error}"
             })
         else:
-            # Format output: preserve line breaks for markdown, remove color codes
-            # Remove ANSI color codes
-            output_clean = re.sub(r'\x1b\[[0-9;]*m', '', output) if output else ""
-            # Remove HTML color tags
-            output_clean = re.sub(r'<span[^>]*style=["\'][^"\']*color[^"\']*["\'][^>]*>|</span>', '', output_clean)
-            # Preserve line breaks with markdown format (two spaces + newline)
-            formatted_output = output_clean.replace('\n', '  \n') if output_clean else ""
+            # Format output: preserve line breaks for markdown, remove color codes and HTML
+            if output:
+                # Remove ANSI color codes
+                output_clean = re.sub(r'\x1b\[[0-9;]*m', '', output)
+                # Remove HTML tags with style attributes
+                output_clean = re.sub(r'<[^>]*style=[^>]*>', '', output_clean)
+                # Remove closing tags
+                output_clean = re.sub(r'</[^>]+>', '', output_clean)
+                # Remove HTML color/font tags
+                output_clean = re.sub(r'<(font|span)[^>]*>', '', output_clean)
+                # Remove any remaining HTML comments
+                output_clean = re.sub(r'<!--.*?-->', '', output_clean, flags=re.DOTALL)
+                # Preserve line breaks with markdown format (two spaces + newline)
+                formatted_output = output_clean.replace('\n', '  \n')
+            else:
+                formatted_output = ""
             results.append({
                 "Model": model_name,
                 "Time (s)": f"{elapsed_time:.2f}",
@@ -776,7 +850,7 @@ def clear_models():
     selected_models.clear()
     return create_model_list_display(), "🗑️ All models cleared"
 
-def export_config(openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key):
+def export_config(openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, kimi_key):
     """Export current model configuration with API keys as JSON"""
     if not selected_models:
         return "", gr.File(visible=False, value=None), "⚠️ No models to export"
@@ -791,6 +865,7 @@ def export_config(openai_key, anthropic_key, google_key, cohere_key, mistral_key
             "Mistral": mistral_key,
             "Deepseek": deepseek_key,
             "Qwen": qwen_key,
+            "Kimi": kimi_key,
         },
         "models": selected_models
     }
@@ -808,13 +883,13 @@ def export_config(openai_key, anthropic_key, google_key, cohere_key, mistral_key
 def import_config(config_text):
     """Import model configuration with API keys from JSON"""
     if not config_text or config_text.strip() == "":
-        return create_model_list_display(), "", "", "", "", "", "", "", "⚠️ Please paste a configuration to import"
+        return create_model_list_display(), "", "", "", "", "", "", "", "", "⚠️ Please paste a configuration to import"
     
     try:
         config = json.loads(config_text)
         
         if "models" not in config:
-            return create_model_list_display(), "", "", "", "", "", "", "", "⚠️ Invalid configuration format"
+            return create_model_list_display(), "", "", "", "", "", "", "", "", "⚠️ Invalid configuration format"
         
         # Clear existing models
         selected_models.clear()
@@ -836,15 +911,16 @@ def import_config(config_text):
         mistral_key = api_keys.get("Mistral", "")
         deepseek_key = api_keys.get("Deepseek", "")
         qwen_key = api_keys.get("Qwen", "")
+        kimi_key = api_keys.get("Kimi", "")
         
         return (create_model_list_display(), 
-                openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key,
+                openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, kimi_key,
                 f"✅ Imported {len(selected_models)} model(s) with API keys")
         
     except json.JSONDecodeError as e:
-        return create_model_list_display(), "", "", "", "", "", "", "", f"⚠️ Invalid JSON format: {str(e)}"
+        return create_model_list_display(), "", "", "", "", "", "", "", "", f"⚠️ Invalid JSON format: {str(e)}"
     except Exception as e:
-        return create_model_list_display(), "", "", "", "", "", "", "", f"⚠️ Import failed: {str(e)}"
+        return create_model_list_display(), "", "", "", "", "", "", "", "", f"⚠️ Import failed: {str(e)}"
 
 # Global state for panel visibility
 panel_visible = True
@@ -858,6 +934,33 @@ def toggle_config_panel():
     else:
         return gr.Column(visible=False), gr.Button("▶ SHOW CONFIG")
 
+# Global state for fullscreen
+fullscreen_mode = False
+results_data = None
+
+def toggle_fullscreen_results():
+    """Toggle fullscreen mode for results - only hide config"""
+    global fullscreen_mode
+    fullscreen_mode = not fullscreen_mode
+    if fullscreen_mode:
+        return gr.Column(visible=False), gr.Button("⛶ EXIT FULLSCREEN", variant="secondary", size="sm")
+    else:
+        return gr.Column(visible=True), gr.Button("⛶ FULLSCREEN", variant="secondary", size="sm")
+
+def filter_results(dataframe, hide_errors):
+    """Filter results to hide error rows if checkbox is checked"""
+    if dataframe is None or dataframe.empty:
+        return dataframe
+    
+    if hide_errors:
+        # Filter out rows where Time or Cost column contains "ERROR"
+        filtered_df = dataframe[
+            (dataframe.iloc[:, 1] != "ERROR") & 
+            (dataframe.iloc[:, 2] != "ERROR")
+        ]
+        return filtered_df
+    return dataframe
+
 # Create clean Matrix-style theme
 matrix_theme = gr.themes.Base(
     primary_hue="green",
@@ -890,6 +993,53 @@ with gr.Blocks(title="Multi-LLM Comparison Tool", theme=matrix_theme, css="""
     .gr-box {border: none !important;}
     .gr-form {border: none !important; box-shadow: none !important;}
     .gr-group {border: none !important;}
+    
+    /* Clean markdown rendering in dataframe */
+    .dataframe .markdown {
+        color: #c9d1d9 !important;
+        font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Helvetica, Arial, sans-serif;
+        line-height: 1.6;
+    }
+    .dataframe .markdown h1, 
+    .dataframe .markdown h2, 
+    .dataframe .markdown h3,
+    .dataframe .markdown h4,
+    .dataframe .markdown h5,
+    .dataframe .markdown h6 {
+        color: #c9d1d9 !important;
+        font-weight: 600;
+        margin-top: 0.5em;
+        margin-bottom: 0.5em;
+    }
+    .dataframe .markdown strong {
+        font-weight: 700;
+        color: #c9d1d9 !important;
+    }
+    .dataframe .markdown em {
+        font-style: italic;
+        color: #c9d1d9 !important;
+    }
+    .dataframe .markdown code {
+        background: #161b22;
+        padding: 0.2em 0.4em;
+        border-radius: 3px;
+        font-family: monospace;
+        color: #c9d1d9 !important;
+    }
+    .dataframe .markdown pre {
+        background: #161b22;
+        padding: 10px;
+        border-radius: 5px;
+        overflow-x: auto;
+    }
+    .dataframe .markdown ul, .dataframe .markdown ol {
+        padding-left: 2em;
+        color: #c9d1d9 !important;
+    }
+    .dataframe .markdown a {
+        color: #58a6ff !important;
+        text-decoration: none;
+    }
 """) as demo:
     
     gr.Markdown("# ⚡ MULTI-LLM COMPARISON MATRIX ⚡")
@@ -973,6 +1123,14 @@ with gr.Blocks(title="Multi-LLM Comparison Tool", theme=matrix_theme, css="""
                 info="Get key: https://dashscope.console.aliyun.com/"
             )
             
+            kimi_key = gr.Textbox(
+                label="🔑 Kimi API Key", 
+                type="password", 
+                placeholder="sk-...", 
+                visible=True,
+                info="Get key: https://platform.moonshot.cn/"
+            )
+            
             with gr.Row():
                 input_price = gr.Number(value=0.15, label="💰 Input $/1M", precision=2, minimum=0, scale=1)
                 output_price = gr.Number(value=0.60, label="Output $/1M", precision=2, minimum=0, scale=1)
@@ -982,7 +1140,7 @@ with gr.Blocks(title="Multi-LLM Comparison Tool", theme=matrix_theme, css="""
                 top_p_slider = gr.Number(value=1.0, label="Top-P", minimum=0, maximum=1, step=0.05, scale=1)
             
             with gr.Row():
-                max_tokens_slider = gr.Number(value=1000, label="Max Tokens", minimum=1, maximum=4096, step=1, scale=2)
+                max_tokens_slider = gr.Number(value=2000, label="Max Tokens", minimum=1, maximum=8192, step=1, scale=2)
                 top_k_number = gr.Number(value=0, label="Top-K", minimum=0, step=1, visible=True, scale=1)
             
             with gr.Row():
@@ -1014,7 +1172,8 @@ with gr.Blocks(title="Multi-LLM Comparison Tool", theme=matrix_theme, css="""
             
             import_status = gr.Markdown("")
         
-        with gr.Column(scale=2):
+        prompt_column = gr.Column(scale=2, visible=True)
+        with prompt_column:
             gr.Markdown("### 💬 INPUT PROMPT")
             prompt_input = gr.Textbox(
                 label="",
@@ -1027,13 +1186,18 @@ with gr.Blocks(title="Multi-LLM Comparison Tool", theme=matrix_theme, css="""
             
             comparison_status = gr.Markdown("")
             
-            gr.Markdown("### 📊 RESULTS MATRIX")
+            with gr.Row():
+                gr.Markdown("### 📊 RESULTS MATRIX")
+                with gr.Row():
+                    hide_errors_checkbox = gr.Checkbox(label="Hide Errors", value=False, scale=1)
+                    fullscreen_results_btn = gr.Button("⛶ FULLSCREEN", variant="secondary", size="sm", scale=1)
+            
             results_table = gr.Dataframe(
                 headers=["Model", "Time (s)", "Est. Cost per 1000 calls ($)", "Output"],
                 wrap=True,
                 interactive=False,
-                datatype=["str", "str", "str", "markdown"],
-                column_widths=["20%", "10%", "15%", "55%"]
+                datatype=["markdown", "str", "str", "markdown"],
+                column_widths=["25%", "10%", "15%", "50%"]
             )
             
             export_btn = gr.Button("📥 DOWNLOAD AS CSV", variant="secondary", visible=False)
@@ -1043,7 +1207,7 @@ with gr.Blocks(title="Multi-LLM Comparison Tool", theme=matrix_theme, css="""
     provider_dropdown.change(
         fn=update_on_provider_change,
         inputs=[provider_dropdown],
-        outputs=[model_dropdown, openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, input_price, output_price]
+        outputs=[model_dropdown, openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, kimi_key, input_price, output_price]
     )
     
     model_dropdown.change(
@@ -1063,7 +1227,7 @@ with gr.Blocks(title="Multi-LLM Comparison Tool", theme=matrix_theme, css="""
         inputs=[provider_dropdown, model_dropdown, temperature_slider, top_p_slider, 
                 max_tokens_slider, top_k_number, freq_penalty_slider, pres_penalty_slider,
                 input_price, output_price, openai_key, anthropic_key, google_key, 
-                cohere_key, mistral_key, deepseek_key, qwen_key],
+                cohere_key, mistral_key, deepseek_key, qwen_key, kimi_key],
         outputs=[models_display, status_text]
     )
     
@@ -1074,7 +1238,7 @@ with gr.Blocks(title="Multi-LLM Comparison Tool", theme=matrix_theme, css="""
     
     run_btn.click(
         fn=run_comparison,
-        inputs=[prompt_input, openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key],
+        inputs=[prompt_input, openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, kimi_key],
         outputs=[results_table, comparison_status, export_btn]
     )
     
@@ -1089,20 +1253,31 @@ with gr.Blocks(title="Multi-LLM Comparison Tool", theme=matrix_theme, css="""
     
     export_config_btn.click(
         fn=export_config,
-        inputs=[openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key],
+        inputs=[openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, kimi_key],
         outputs=[import_textbox, config_download_file, import_status]
     )
     
     import_btn.click(
         fn=import_config,
         inputs=[import_textbox],
-        outputs=[models_display, openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, import_status]
+        outputs=[models_display, openai_key, anthropic_key, google_key, cohere_key, mistral_key, deepseek_key, qwen_key, kimi_key, import_status]
     )
     
     toggle_panel_btn.click(
         fn=toggle_config_panel,
         outputs=[config_column, toggle_panel_btn]
     )
+    
+    fullscreen_results_btn.click(
+        fn=toggle_fullscreen_results,
+        outputs=[config_column, fullscreen_results_btn]
+    )
+    
+    hide_errors_checkbox.change(
+        fn=filter_results,
+        inputs=[results_table, hide_errors_checkbox],
+        outputs=[results_table]
+    )
 
 if __name__ == "__main__":
     demo.launch(share=True)