PILL

app.py

@@ -44,7 +44,7 @@ def provider_change(provider, selected_type, all_types=["text", "vision", "video
     return new_models if new_models else all_models
 
 # --------------------------
-# Estimate Cost Function (Updated)
+# Estimate Cost Function
 # --------------------------
 def estimate_cost(num_alerts, input_size, output_size, model_id):
     pricing = st.session_state.get("pricing", {})
@@ -80,35 +80,68 @@ if "data_loaded" not in st.session_state:
 with st.sidebar:
     st.image("https://cdn.prod.website-files.com/630f558f2a15ca1e88a2f774/631f1436ad7a0605fecc5e15_Logo.svg",
              use_container_width=True)
-    st.markdown(
-        """ Visit: [https://www.priam.ai](https://www.priam.ai)
-        """
-    )
+    st.markdown("Visit: [https://www.priam.ai](https://www.priam.ai)")
     st.divider()
     st.sidebar.title("LLM Pricing Calculator")
 
-# Track active tab in session state
-if "active_tab" not in st.session_state:
-    st.session_state.active_tab = "Model Selection"
-
-def switch_tab(tab_name):
-    st.session_state.active_tab = tab_name
-    st.rerun()
-
-
 # --------------------------
-# Main Content Layout (Tabs)
+# Pills Navigation (Using st.pills)
 # --------------------------
-tab_labels = ["Model Selection", "On Premise Estimator", "About"]
-tab_index = tab_labels.index(st.session_state.active_tab)
-tabs = st.tabs(tab_labels)
-tab1, tab2, tab3 = tabs
+# st.pills creates a pill-style selection widget.
+page = st.pills("Head",
+    options=["Model Selection", "On Premise Estimator", "About"],selection_mode="single",default="Model Selection",label_visibility="hidden",
+    #index=0  # Change index if you want a different default
+)
 
+# --------------------------
+# Helper: Format Analysis Report
+# --------------------------
+def format_analysis_report(analysis_result: dict) -> str:
+    """Convert the raw analysis_result dict into a human-readable report."""
+    if "error" in analysis_result:
+        return f"**Error:** {analysis_result['error']}"
+    
+    lines = []
+    lines.append(f"### Model Analysis Report for `{analysis_result.get('model_id', 'Unknown Model')}`\n")
+    lines.append(f"**Parameter Size:** {analysis_result.get('parameter_size', 'N/A')} Billion parameters\n")
+    lines.append(f"**Precision:** {analysis_result.get('precision', 'N/A')}\n")
+    
+    vram = analysis_result.get("vram_requirements", {})
+    lines.append("#### VRAM Requirements:")
+    lines.append(f"- Model Size: {vram.get('model_size_gb', 0):.2f} GB")
+    lines.append(f"- KV Cache: {vram.get('kv_cache_gb', 0):.2f} GB")
+    lines.append(f"- Activations: {vram.get('activations_gb', 0):.2f} GB")
+    lines.append(f"- Overhead: {vram.get('overhead_gb', 0):.2f} GB")
+    lines.append(f"- **Total VRAM:** {vram.get('total_vram_gb', 0):.2f} GB\n")
+    
+    compatible_gpus = analysis_result.get("compatible_gpus", [])
+    lines.append("#### Compatible GPUs:")
+    if compatible_gpus:
+        for gpu in compatible_gpus:
+            lines.append(f"- {gpu}")
+    else:
+        lines.append("- None found")
+    lines.append(f"\n**Largest Compatible GPU:** {analysis_result.get('largest_compatible_gpu', 'N/A')}\n")
+    
+    gpu_perf = analysis_result.get("gpu_performance", {})
+    if gpu_perf:
+        lines.append("#### GPU Performance:")
+        for gpu, perf in gpu_perf.items():
+            lines.append(f"**{gpu}:**")
+            lines.append(f"  - Tokens per Second: {perf.get('tokens_per_second', 0):.2f}")
+            lines.append(f"  - FLOPs per Token: {perf.get('flops_per_token', 0):.2f}")
+            lines.append(f"  - Effective TFLOPS: {perf.get('effective_tflops', 0):.2f}\n")
+    else:
+        lines.append("#### GPU Performance: N/A\n")
+    
+    return "\n".join(lines)
 
-# ----- Tab 1: Model Selection -----
-with tab1:
+# --------------------------
+# Render Content Based on Selected Pill
+# --------------------------
+if page == "Model Selection":
+    st.divider()
     st.header("LLM Pricing App")
-
     # --- Row 1: Provider/Type and Model Selection ---
     col_left, col_right = st.columns(2)
     with col_left:
@@ -118,50 +151,27 @@ with tab1:
             index=st.session_state["providers"].index("azure") if "azure" in st.session_state["providers"] else 0
         )
         selected_type = st.radio("Select type", options=["text", "image"], index=0)
-
     with col_right:
-        # Filter models based on the selected provider and type
         filtered_models = provider_change(selected_provider, selected_type)
-        
         if filtered_models:
-            # Force "gpt-4-turbo" as default if available; otherwise, default to the first model.
             default_model = "o1" if "o1" in filtered_models else filtered_models[0]
-            selected_model = st.selectbox(
-                "Select a model",
-                options=filtered_models,
-                index=filtered_models.index(default_model)
-            )
+            selected_model = st.selectbox("Select a model", options=filtered_models, index=filtered_models.index(default_model))
         else:
             selected_model = None
             st.write("No models available")
-
+    
     # --- Row 2: Alert Stats ---
     col1, col2, col3 = st.columns(3)
     with col1:
-        num_alerts = st.number_input(
-            "Security Alerts Per Day",
-            value=100,
-            min_value=1,
-            step=1,
-            help="Number of security alerts to analyze daily"
-        )
+        num_alerts = st.number_input("Security Alerts Per Day", value=100, min_value=1, step=1,
+                                     help="Number of security alerts to analyze daily")
     with col2:
-        input_size = st.number_input(
-            "Alert Content Size (characters)",
-            value=1000,
-            min_value=1,
-            step=1,
-            help="Include logs, metadata, and context per alert"
-        )
+        input_size = st.number_input("Alert Content Size (characters)", value=1000, min_value=1, step=1,
+                                     help="Include logs, metadata, and context per alert")
     with col3:
-        output_size = st.number_input(
-            "Analysis Output Size (characters)",
-            value=500,
-            min_value=1,
-            step=1,
-            help="Expected length of security analysis and recommendations"
-        )
-
+        output_size = st.number_input("Analysis Output Size (characters)", value=500, min_value=1, step=1,
+                                      help="Expected length of security analysis and recommendations")
+    
     # --- Row 3: Buttons ---
     btn_col1, btn_col2 = st.columns(2)
     with btn_col1:
@@ -178,99 +188,34 @@ with tab1:
                 st.session_state["pricing"] = pricing
                 st.session_state["providers"] = providers
                 st.success("Pricing data refreshed!")
-
+    
     st.divider()
-    # --- Display Results ---
     st.markdown("### Results")
     if "result" in st.session_state:
         st.write(st.session_state["result"])
     else:
         st.write("Use the buttons above to estimate costs.")
-
-    # --- Clear Button Below Results ---
+    
     if st.button("Clear"):
         st.session_state.pop("result", None)
-        st.rerun()
-
-
-# ----- Tab 2: On Premise Estimator -----
-def format_analysis_report(analysis_result: dict) -> str:
-    """Convert the raw analysis_result dict into a human-readable report."""
-    if "error" in analysis_result:
-        return f"**Error:** {analysis_result['error']}"
-    
-    lines = []
-    lines.append(f"### Model Analysis Report for `{analysis_result.get('model_id', 'Unknown Model')}`\n")
-    lines.append(f"**Parameter Size:** {analysis_result.get('parameter_size', 'N/A')} Billion parameters\n")
-    lines.append(f"**Precision:** {analysis_result.get('precision', 'N/A')}\n")
-    
-    vram = analysis_result.get("vram_requirements", {})
-    lines.append("#### VRAM Requirements:")
-    lines.append(f"- Model Size: {vram.get('model_size_gb', 0):.2f} GB")
-    lines.append(f"- KV Cache: {vram.get('kv_cache_gb', 0):.2f} GB")
-    lines.append(f"- Activations: {vram.get('activations_gb', 0):.2f} GB")
-    lines.append(f"- Overhead: {vram.get('overhead_gb', 0):.2f} GB")
-    lines.append(f"- **Total VRAM:** {vram.get('total_vram_gb', 0):.2f} GB\n")
-    
-    compatible_gpus = analysis_result.get("compatible_gpus", [])
-    lines.append("#### Compatible GPUs:")
-    if compatible_gpus:
-        for gpu in compatible_gpus:
-            lines.append(f"- {gpu}")
-    else:
-        lines.append("- None found")
-    lines.append(f"\n**Largest Compatible GPU:** {analysis_result.get('largest_compatible_gpu', 'N/A')}\n")
-    
-    gpu_perf = analysis_result.get("gpu_performance", {})
-    if gpu_perf:
-        lines.append("#### GPU Performance:")
-        for gpu, perf in gpu_perf.items():
-            lines.append(f"**{gpu}:**")
-            lines.append(f"  - Tokens per Second: {perf.get('tokens_per_second', 0):.2f}")
-            lines.append(f"  - FLOPs per Token: {perf.get('flops_per_token', 0):.2f}")
-            lines.append(f"  - Effective TFLOPS: {perf.get('effective_tflops', 0):.2f}\n")
-    else:
-        lines.append("#### GPU Performance: N/A\n")
-   
-    #model_info = analysis_result.get("model_info", {})
-    #lines.append("#### Model Information:")
-    #if model_info:
-     #   if model_info.get("description"):
-      #      lines.append(f"- Description: {model_info['description']}")
-       # if model_info.get("tags"):
-        #    lines.append(f"- Tags: {', '.join(model_info['tags'])}")
-        #if model_info.get("downloads") is not None:
-         #   lines.append(f"- Downloads: {model_info['downloads']}")
-        #if model_info.get("library"):
-         #   lines.append(f"- Library: {model_info['library']}")
-    #else:
-     #   lines.append("No additional model info available.")
-    
-    return "\n".join(lines)
-
 
-# ----- Tab 2: On Premise Estimator -----
-with tab2:
+elif page == "On Premise Estimator":
+    st.divider()
     st.header("On Premise Estimator")
     st.markdown("Enter a Hugging Face model ID to perform an on premise analysis using the provided estimator.")
-
-    # Input for model ID with a default value
     hf_model_id = st.text_input("Hugging Face Model ID", value="facebook/opt-1.3b")
     
     if st.button("Analyze Model"):
-        st.session_state.active_tab = "On Premise Estimator"
         with st.spinner("Analyzing model..."):
             analysis_result = analyze_hf_model(hf_model_id)
-        st.session_state.analysis_result = analysis_result
-        st.rerun()
-
-    # Render if analysis result exists
+        st.session_state["analysis_result"] = analysis_result
+    
     if "analysis_result" in st.session_state:
-        report = format_analysis_report(st.session_state.analysis_result)
+        report = format_analysis_report(st.session_state["analysis_result"])
         st.markdown(report)
 
-# ----- Tab 3: About -----
-with tab3:
+elif page == "About":
+    st.divider()
     st.markdown(
         """
         ## About This App
@@ -279,7 +224,7 @@ with tab3:
 
         - The app downloads the latest pricing from the LiteLLM repository.
         - Using simple maths to estimate the total tokens.
-        - helps you estimate hardware requirements for running open-source large language models (LLMs) on-premise using only the model ID from Hugging Face.
+        - Helps you estimate hardware requirements for running open-source large language models (LLMs) on-premise using only the model ID from Hugging Face.
         - Version 0.1
 
         ---
@@ -288,12 +233,10 @@ with tab3:
 
         | Version | Release Date | Key Feature Updates |
         |--------|--------------|---------------------|
-        | `v1.0` | 2025-03-26  | Initial release with basic total tokens estimation|
-        | `v1.1` | 2025-04-06   | Added On premise Estimator Tab |
-
+        | `v1.0` | 2025-03-26  | Initial release with basic total tokens estimation |
+        | `v1.1` | 2025-04-06  | Added On Premise Estimator Tab |
 
         ---
-    
 
         Website: [https://www.priam.ai](https://www.priam.ai)
         """
@@ -304,4 +247,4 @@ with tab3:
 
         This app is for demonstration purposes only. Actual costs may vary based on usage patterns and other factors.
         """
-    )
+    )