Create app.py

app.py

@@ -0,0 +1,140 @@
+import re
+import streamlit as st
+from together import Together
+
+# Initialize the page
+st.set_page_config(page_title="AI Prompt Evaluator", page_icon="🧠", layout="wide")
+st.title("AI Prompt Evaluator")
+
+def evaluate_prompt(prompt):
+    """Evaluates a prompt based on key principles of good prompt engineering."""
+    
+    criteria = {
+        "clarity": bool(re.search(r"\b(who|what|where|when|why|how)\b", prompt.lower())),
+        "specificity": len(prompt.split()) > 5,
+        "context": bool(re.search(r"\b(for example|such as|like|including)\b", prompt.lower())),
+        "output_format": bool(re.search(r"\b(list|table|bullets|code|summary)\b", prompt.lower())),
+        "constraints": bool(re.search(r"\b(limit|max|min|exactly|within)\b", prompt.lower()))
+    }
+    
+    score = sum(criteria.values()) * 2  # Scale to 10 points
+    
+    suggestions = []
+    if not criteria["clarity"]:
+        suggestions.append("Make the prompt clearer by specifying exactly what you need.")
+    if not criteria["specificity"]:
+        suggestions.append("Make the prompt more detailed and specific.")
+    if not criteria["context"]:
+        suggestions.append("Add some background information to improve relevance.")
+    if not criteria["output_format"]:
+        suggestions.append("Specify how you want the output to be structured (e.g., list, code).")
+    if not criteria["constraints"]:
+        suggestions.append("Define any limits or conditions (e.g., word count, time frame).")
+    
+    return score, suggestions, criteria
+
+def get_ai_response(prompt, model):
+    """Get a response from the Together API"""
+    try:
+        client = Together(api_key=st.secrets["together_api_key"])
+        response = client.chat.completions.create(
+            model=model,
+            messages=[{"role": "user", "content": prompt}],
+            max_tokens=500
+        )
+        return response.choices[0].message.content
+    except Exception as e:
+        return f"Error: {str(e)}"
+
+# Sidebar for configuration
+st.sidebar.header("Configuration")
+model_options = {
+    "Mistral 7B": "mistralai/Mistral-7B-Instruct-v0.3",
+    "Llama 3 8B": "meta-llama/Llama-3-8B-Instruct",
+    "Qwen 72B": "Qwen/Qwen-72B-Chat"
+}
+selected_model = st.sidebar.selectbox("Select AI Model", list(model_options.keys()))
+
+# Main input area
+prompt_input = st.text_area("Enter your prompt:", height=150)
+
+col1, col2 = st.columns(2)
+with col1:
+    if st.button("Evaluate Prompt", type="primary"):
+        if not prompt_input.strip():
+            st.error("Please enter a prompt to evaluate.")
+        else:
+            # Store API key in secrets
+            if "together_api_key" not in st.secrets:
+                st.secrets["together_api_key"] = "4db152889da5afebdba262f90e4cdcf12976ee8b48d9135c2bb86ef9b0d12bdd"
+            
+            score, suggestions, criteria = evaluate_prompt(prompt_input)
+            
+            # Display evaluation results
+            st.subheader("Prompt Evaluation Results")
+            
+            # Score with color-coded meter
+            if score >= 8:
+                st.success(f"Score: {score}/10 - Excellent!")
+            elif score >= 6:
+                st.info(f"Score: {score}/10 - Good")
+            else:
+                st.warning(f"Score: {score}/10 - Needs Improvement")
+            
+            # Criteria checklist
+            st.markdown("#### Criteria Checklist")
+            for criterion, passed in criteria.items():
+                icon = "✅" if passed else "❌"
+                st.markdown(f"{icon} **{criterion.capitalize()}**")
+            
+            # Improvement suggestions
+            if suggestions:
+                st.markdown("#### Suggestions to improve your prompt:")
+                for tip in suggestions:
+                    st.markdown(f"- {tip}")
+            else:
+                st.success("Your prompt is well-structured!")
+            
+            # Get AI response if score is high enough
+            if score >= 4:  # Only get response if score is decent
+                with st.spinner("Getting AI response..."):
+                    ai_response = get_ai_response(prompt_input, model_options[selected_model])
+                    
+                st.subheader("AI Response Preview")
+                st.markdown(ai_response)
+            else:
+                st.info("Improve your prompt score to see an AI response preview.")
+
+with col2:
+    st.subheader("Prompt Engineering Tips")
+    st.markdown("""
+    ### How to craft effective prompts:
+    
+    1. **Be clear and specific** - Clearly state what you want.
+    2. **Provide context** - Give background information.
+    3. **Specify output format** - Ask for lists, code, or summaries.
+    4. **Set constraints** - Define limits like word count.
+    5. **Use examples** - Show examples of desired output.
+    
+    ### Examples of good prompts:
+    
+    - "Create a 5-item bulleted list of healthy breakfast ideas including nutritional benefits."
+    - "Write a Python function that sorts a list of integers using the bubble sort algorithm."
+    - "Summarize the key benefits of exercise in exactly 100 words for a health newsletter."
+    """)
+
+# Add explanation of the evaluation criteria
+st.markdown("---")
+with st.expander("Understanding the Evaluation Criteria"):
+    st.markdown("""
+    - **Clarity**: Does the prompt include clear question words (who, what, where, when, why, how)?
+    - **Specificity**: Is the prompt detailed enough (more than 5 words)?
+    - **Context**: Does the prompt provide examples or background information?
+    - **Output Format**: Does the prompt specify the desired format (list, table, code, etc.)?
+    - **Constraints**: Does the prompt include specific limitations or requirements?
+    """)
+
+# Footer
+st.markdown("---")
+st.markdown("#### About")
+st.markdown("This tool helps you create better prompts for AI systems by evaluating them against best practices in prompt engineering.")