Create app.py

app.py

@@ -0,0 +1,114 @@
+# app.py
+import streamlit as st
+import requests
+import os
+import re
+
+# Environment Setup
+HF_API_URL = "https://router.huggingface.co/novita/v3/openai/chat/completions"
+HF_TOKEN = os.environ.get("HF_PROJECT_TOKEN")
+HEADERS = {"Authorization": f"Bearer {HF_TOKEN}"}
+
+# Prompt Template
+PROMPT_TEMPLATE = """
+You are a prompt evaluation assistant called \"PromptPolice\". Evaluate the following user prompt based on the criteria below. For each, rate from 1 (poor) to 5 (excellent), explain why, and suggest specific improvements.
+
+Prompt:
+\"\"\" 
+{user_prompt} 
+\"\"\"
+
+Evaluation Criteria (1–5):
+- 🔍 Clarity: Is the prompt easy to understand?
+- 🧠 Context: Does the prompt provide enough background or situational framing?
+- 🎯 Specificity: Are the goals or constraints clearly defined?
+- 🛍️ Intent Alignment: Is it clear what task or behavior the prompt is meant to elicit?
+
+Verdict Logic:
+- ✅ **Pass**: Overall Score is 16 or above AND no criterion rated below 3.
+- ⚠️ **Warning**: Score between 11–15 OR one rating below 3.
+- 🚫 **Ticket**: Score 10 or below OR two or more ratings below 3.
+
+Output Format:
+
+Evaluation:
+- 🔍 Clarity: X/5 — [Explanation]
+- 🧠 Context: X/5 — [Explanation]
+- 🎯 Specificity: X/5 — [Explanation]
+- 🛏️ Intent Alignment: X/5 — [Explanation]
+
+Overall Score: X/20
+
+Flaw Summary: [One-line summary of the weakest point]
+
+🛡️ Verdict: [✅ Pass / ⚠️ Warning / 🚫 Ticket] — [Justification based on score and flaw]
+
+Prompt Type(s): [e.g., Instruction, Summarization, Query, Roleplay, Classification, Creative Writing, Other]
+
+Suggestions:
+- [Actionable suggestion 1]
+- [Actionable suggestion 2]
+
+Improved Prompt:
+"[Rewritten version of the user prompt]"
+"""
+
+# Function to query DeepSeek
+@st.cache_data(show_spinner=False)
+def evaluate_prompt(user_prompt):
+    payload = {
+        "messages": [
+            {"role": "user", "content": PROMPT_TEMPLATE.format(user_prompt=user_prompt)}
+        ],
+        "model": "deepseek/deepseek-r1-turbo"
+    }
+    response = requests.post(HF_API_URL, headers=HEADERS, json=payload)
+    if response.status_code == 200:
+        return response.json()["choices"][0]["message"]["content"]
+    else:
+        return f"Error: {response.status_code} - {response.text}"
+
+# App UI
+st.set_page_config("PromptPolice", page_icon="🚓", layout="wide")
+st.title(":oncoming_police_car: PromptPolice: Evaluate Your Prompts Like a Pro")
+
+# Sidebar
+with st.sidebar:
+    st.header(":gear: Prompt Options")
+    use_example = st.checkbox("Load Example Prompt")
+    st.markdown("---")
+    st.write("Created with ❤️ using DeepSeek + Streamlit")
+
+# Main Input
+if use_example:
+    user_input = st.text_area("Paste your prompt here:",
+        "Generate a short story about a robot in a post-apocalyptic world.",
+        height=200)
+else:
+    user_input = st.text_area("Paste your prompt here:", height=200)
+
+# Evaluate Button
+if st.button(":mag_right: Evaluate Prompt"):
+    if not HF_TOKEN:
+        st.error("Missing Hugging Face token. Please set HF_PROJECT_TOKEN as environment variable.")
+    elif user_input.strip() == "":
+        st.warning("Please enter a prompt to evaluate.")
+    else:
+        with st.spinner("Evaluating prompt with PromptPolice..."):
+            result = evaluate_prompt(user_input)
+        st.markdown("---")
+        st.subheader(":clipboard: Evaluation Result")
+        st.markdown(f"""
+        <div style='background-color:#f5f5f5; padding:20px; border-radius:10px;'>
+        <pre style='white-space:pre-wrap'>{result}</pre>
+        </div>
+        """, unsafe_allow_html=True)
+
+        # Option to Copy
+        st.download_button("Download Evaluation", result, file_name="evaluation.txt")
+
+# Footer
+st.markdown("""
+---
+Made with 🌟 by [Penguins]. Powered by DeepSeek R1 Turbo (free-tier). Feedback welcome!
+""")