used tf-idf similarity

app.py

@@ -1,45 +1,49 @@
 import gradio as gr
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.metrics.pairwise import cosine_similarity
+
+# Pre-existing emails
+pre_existing_emails = [
+    "Build system failed due to missing dependencies.",
+    "Agent deployment was successful on the new cluster.",
+    "Cluster performance metrics are above expectations.",
+    "General inquiry about the project timeline.",
+    "Build logs indicate a configuration error.",
+    "Agent is not responding after the last update.",
+    "Cluster nodes need to be restarted.",
+    "General feedback on the new UI.",
+    "Build completed successfully.",
+    "Agent registration failed due to timeout."
+]
+
+# Fit TF-IDF vectorizer on pre-existing emails
+vectorizer = TfidfVectorizer().fit(pre_existing_emails)
+email_vectors = vectorizer.transform(pre_existing_emails)
+
+def semantic_search(query: str) -> dict:
+    # Vectorize the query
+    query_vec = vectorizer.transform([query])
+    # Compute cosine similarity
+    similarities = cosine_similarity(query_vec, email_vectors).flatten()
+    # Get indices of top 5 similar emails
+    top_indices = similarities.argsort()[-5:][::-1]
+    # Prepare results
+    results = [
+        {
+            "email": pre_existing_emails[i],
+            "similarity": float(similarities[i])
+        }
+        for i in top_indices
+    ]
+    return {"results": results}
 
-# Simulated local storage for clustered emails
-email_clusters = {}
-
-def cluster_email(text: str) -> dict:
-    """
-    Assign a topic label to the email and store it in a local cluster dictionary.
-
-    Args:
-        text (str): The email content
-
-    Returns:
-        dict: Topic label and current cluster state
-    """
-    # Simple keyword-based topic detection
-    if "build" in text.lower():
-        topic = "Build"
-    elif "agent" in text.lower():
-        topic = "Agents"
-    elif "cluster" in text.lower():
-        topic = "Clustering"
-    else:
-        topic = "General"
-
-    # Store in cluster
-    email_clusters.setdefault(topic, []).append(text)
-
-    return {
-        "assigned_topic": topic,
-        "cluster_size": len(email_clusters[topic]),
-        "all_topics": list(email_clusters.keys())
-    }
-
-# Gradio interface
 demo = gr.Interface(
-    fn=cluster_email,
-    inputs=gr.Textbox(placeholder="Paste email content here..."),
+    fn=semantic_search,
+    inputs=gr.Textbox(placeholder="Enter your email or query..."),
     outputs=gr.JSON(),
-    title="Email Topic Classifier",
-    description="Assigns a topic to an email and stores it in a local cluster"
+    title="Email Semantic Search",
+    description="Returns the top 5 most similar pre-existing emails based on your query."
 )
 
 if __name__ == "__main__":
-    demo.launch(mcp_server=True)
+    demo.launch(mcp_server=True)

requirements.txt

@@ -1,2 +1,3 @@
 gradio[mcp]
-textblob
+textblob
+scikit-learn