src/run2.py

import streamlit as st

def run_app2():
    st.title("LLM Compatibility Advisor (Manual Spec Entry)")
    st.markdown("Enter your CPU and GPU specs to get local LLM suggestions (Ollama-compatible)")

    st.markdown("### 🔧 How to Find Your Device Configuration")
    st.markdown("""
    - **🪟 Windows**: Press `Win + R`, type `dxdiag`, and press Enter to view CPU, GPU, and RAM details.
    - **🍎 macOS**: Click Apple  menu → `About This Mac` → `More Info...`
    - **🐧 Linux**: Use terminal commands like `lscpu`, `free -h`, or `neofetch`
    """)

    st.markdown("#### 💡 Tip: You can use these tools to help find your processor and GPU names.")

    cpu_name = st.text_input("🖥️ Enter your Processor Name", placeholder="e.g., Intel Core i5-10300H")
    st.caption("ℹ️ Enter your exact CPU name to help identify its capability (found in system settings).")

    with st.expander("Suggested Processor Examples"):
        st.markdown("""
        - **Intel i3**: i3-10100U, i3-1115G4, i3-N305  
        - **Intel i5**: i5-10300H, i5-1240P, i5-13400  
        - **Intel i7**: i7-9750H, i7-11800H, i7-13700K  
        - **Apple M Series**: M1, M1 Pro, M2, M3  
        - **AMD Ryzen**: Ryzen 5 5600H, Ryzen 7 5800X, Ryzen 9 7945HX  
        - **Snapdragon**: Snapdragon 8 Gen 1, Snapdragon 7c Gen 2
        """)

    cpu_category = st.selectbox(
        "⚙️ Select your CPU Category",
        ["Intel i3", "Intel i5", "Intel i7", "Apple M Series", "AMD Ryzen", "Qualcomm Snapdragon", "Other/Unknown"]
    )
    st.caption("ℹ️ If unsure, choose the closest matching category from the dropdown.")

    gpu_name = st.text_input("🎮 GPU (Optional)", placeholder="e.g., NVIDIA GTX 1650 / None")
    st.caption("ℹ️ GPU helps speed up models. If you're unsure or using only CPU, leave it blank or type 'None'.")

    performance_score = {
        "Intel i3": ("🔴 Low-end", "May only support 4-bit quantized models."),
        "Intel i5": ("🟡 Moderate", "Can run most 4-bit and some 8-bit models."),
        "Intel i7": ("🟢 High-end", "Handles 8-bit and some full FP16 models."),
        "Apple M Series": ("🟢 High-end", "Great efficiency for quantized models."),
        "AMD Ryzen": ("🟢 High-end", "Multi-core power suitable for larger models."),
        "Qualcomm Snapdragon": ("🔴 Low-end", "Best for smallest on-device models."),
        "Other/Unknown": ("🟡 Average", "Limited info—may vary by chip.")
    }

    llm_recommendations = {
        "Intel i3": {
            "Coding": ["Code Llama (7B - quantized)", "Phi-2"],
            "Math & Logic": ["Mistral (7B - quantized)", "Gemma 2B"],
            "General": ["Phi-2", "TinyLlama"]
        },
        "Intel i5": {
            "Coding": ["Code Llama (7B)", "Deepseek Coder (6.7B)"],
            "Math & Logic": ["Mistral 7B", "Gemma 7B"],
            "General": ["Phi-2", "Mistral", "LLaMA 2 (7B)"]
        },
        "Intel i7": {
            "Coding": ["Code Llama (13B - Q4)", "Deepseek Coder 6.7B"],
            "Math & Logic": ["Mistral 7B", "LLaMA 2 13B (quantized)"],
            "General": ["LLaMA 2 (13B)", "OpenChat 3.5"]
        },
        "Apple M Series": {
            "Coding": ["Code Llama 7B (Q4)", "Phi-2"],
            "Math & Logic": ["Gemma 7B", "Mistral (quantized)"],
            "General": ["Mistral", "LLaMA 2 7B", "Phi-2"]
        },
        "AMD Ryzen": {
            "Coding": ["Deepseek Coder", "Code Llama"],
            "Math & Logic": ["Mistral", "LLaMA 2"],
            "General": ["Phi-2", "Mistral", "LLaMA 2"]
        },
        "Qualcomm Snapdragon": {
            "Coding": ["Phi-2 (on-device)"],
            "Math & Logic": ["TinyLlama", "Phi-2"],
            "General": ["TinyLlama", "Gemma 2B"]
        },
        "Other/Unknown": {
            "Coding": ["Phi-2", "TinyLlama"],
            "Math & Logic": ["Gemma 2B", "TinyLlama"],
            "General": ["Phi-2", "TinyLlama"]
        }
    }

    quantized_sizes = {
        "TinyLlama": "FP16: 0.6GB, 8-bit: 0.3GB, 4-bit: 0.15GB",
        "Phi-2": "FP16: 5.2GB, 8-bit: 2.6GB, 4-bit: 1.3GB",
        "Mistral": "FP16: 13GB, 8-bit: 7GB, 4-bit: 3.5GB",
        "Gemma 2B": "FP16: 4.2GB, 8-bit: 2.1GB, 4-bit: 1.1GB",
        "Gemma 7B": "FP16: 13GB, 8-bit: 6.5GB, 4-bit: 3.2GB",
        "Code Llama": "7B: FP16: 13GB, 8-bit: 6.5GB, 4-bit: 3.3GB | 13B: FP16: 26GB, 8-bit: 13GB, 4-bit: 6.5GB",
        "Deepseek Coder": "6.7B: FP16: 12.8GB, 8-bit: 6.4GB, 4-bit: 3.2GB",
        "LLaMA 2": "7B: FP16: 13GB, 8-bit: 6.7GB, 4-bit: 3.5GB | 13B: FP16: 26GB, 8-bit: 13GB, 4-bit: 6.5GB",
        "OpenChat 3.5": "FP16: 7.1GB, 8-bit: 3.6GB, 4-bit: 1.8GB"
    }

    if cpu_name:
        st.markdown("---")
        st.subheader("Your Hardware Configuration")
        st.write(f"**Processor Name:** {cpu_name}")
        st.write(f"**CPU Category:** {cpu_category}")
        st.write(f"**GPU:** {gpu_name or 'Not specified'}")

        score_label, score_note = performance_score.get(cpu_category, ("🟡 Unknown", "Estimate based on general category."))
        st.success(f"📈 Performance Score: {score_label}")
        st.caption(score_note)

        st.markdown("---")
        st.subheader("Recommended LLMs for Local Use (Ollama Compatible)")

        recommendations_text = f"Processor: {cpu_name} ({cpu_category})\\nGPU: {gpu_name or 'None'}\\nPerformance Score: {score_label}\\n\\nRecommended Models:\\n"

        recs = llm_recommendations.get(cpu_category, llm_recommendations["Other/Unknown"])
        for task in ["Coding", "Math & Logic", "General"]:
            st.markdown(f"### 🔹 {task}")
            recommendations_text += f"\\n{task}:\\n"
            for model in recs[task]:
                st.markdown(f"- **{model}**")
                recommendations_text += f"- {model}"
                for key in quantized_sizes:
                    if key.lower() in model.lower():
                        st.caption(f"💾 {quantized_sizes[key]}")
                        recommendations_text += f" ({quantized_sizes[key]})"
                        break
                # Generate command
                cmd = model.split("(")[0].strip().lower().replace(" ", "_").replace("-", "")
                st.code(f"ollama pull {cmd}", language="bash")
                recommendations_text += f"\\n  Command: ollama pull {cmd}\\n"

        st.markdown("---")
        st.markdown(" _Tip: Run these models using `ollama run <model>` or download with `ollama pull <model>`._")

        st.download_button("📥 Download These Recommendations", recommendations_text, file_name="llm_suggestions.txt")
    else:
        st.info("Enter your processor details to see LLM recommendations.")