Merge feature/hf-deployment into master

.github/workflows/push-to-hf-space.yml

@@ -0,0 +1,37 @@
+name: Push to Hugging Face Space
+
+on:
+  push:
+    branches: [ feature/hf-deployment ]   # change if you use another default branch
+
+jobs:
+  sync:
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Check out repo
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0   # keep history; helpful for subtree/sparse etc.
+
+      - name: Set up Git LFS (if you track large files)
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y git-lfs
+          git lfs install
+
+      - name: Push to Space
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+          HF_SPACE: pxdelta/RAG   # or use ${{ secrets.HF_SPACE }}
+        run: |
+          # Configure git identity for the CI push
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+
+          # Add the Space as an authenticated remote.
+          # Username can be anything (often 'oauth2' or your HF username); token goes in the password slot.
+          git remote add hf "https://oauth2:${HF_TOKEN}@huggingface.co/spaces/${HF_SPACE}"
+
+          # Push the current GitHub branch (e.g., main) to the Space's main branch
+          git push hf HEAD:main --force

.gitignore

@@ -2,10 +2,6 @@
 .env
 venv/
 env/
-.venv
-
-notes.txt
-test.ipynb
 
 # Python
 __pycache__/
@@ -22,6 +18,3 @@ Thumbs.db
 # Generated files
 database/university.db
 rag/vector_store/
-
-
-secrets_local.py

Dockerfile

@@ -0,0 +1,29 @@
+# Use the official Python image
+FROM python:3.10-slim
+
+# Create a non-root user
+RUN useradd -m -u 1000 user
+
+# Set the working directory
+WORKDIR /app
+
+# Copy all application files
+COPY . .
+
+# Grant ownership of the app directory to the user
+RUN chown -R user:user /app
+
+# Switch to the non-root user
+USER user
+
+# Install dependencies
+RUN pip install --user -r requirements.txt
+
+# Make the startup script executable
+RUN chmod +x startup.sh
+
+# Add the user's local bin directory to the PATH
+ENV PATH="/home/user/.local/bin:${PATH}"
+
+# Set the entrypoint to the startup script
+CMD ["./startup.sh"]

guards/input.py

@@ -0,0 +1,30 @@
+import re
+
+def is_valid(query: str) -> bool:
+    """
+    Validates the user's query.
+    """
+    # Check for query length
+    if len(query) > 500:
+        return False
+
+    # Check for SQL injection patterns
+    sql_injection_patterns = [
+        r"(\s*(--|#|;))",
+        r"(\s*(union|select|insert|update|delete|drop|alter)\s+)",
+    ]
+    for pattern in sql_injection_patterns:
+        if re.search(pattern, query, re.IGNORECASE):
+            return False
+
+    return True
+
+if __name__ == '__main__':
+    # Example usage
+    valid_query = "What are the computer science courses?"
+    invalid_query_long = "a" * 501
+    invalid_query_sql = "SELECT * FROM students;"
+
+    print(f"'{valid_query}' is valid: {is_valid(valid_query)}")
+    print(f"'long query' is valid: {is_valid(invalid_query_long)}")
+    print(f"'{invalid_query_sql}' is valid: {is_valid(invalid_query_sql)}")

guards/output.py

@@ -0,0 +1,27 @@
+import re
+
+def sanitize(response: str) -> str:
+    """
+    Sanitizes the LLM's response.
+    """
+    # Redact email addresses
+    response = re.sub(r"[a-zA-Z0-9_.+-]+@[a-zA-Z0-9-]+\.[a-zA-Z0-9-.]+", "[REDACTED EMAIL]", response)
+
+    # Check for hallucination patterns
+    hallucination_patterns = [
+        r"as an ai language model",
+        r"i cannot answer that question",
+    ]
+    for pattern in hallucination_patterns:
+        if re.search(pattern, response, re.IGNORECASE):
+            return "I'm sorry, I don't have enough information to answer that question."
+
+    return response
+
+if __name__ == '__main__':
+    # Example usage
+    valid_response = "Professor Smith's email is john.smith@university.edu."
+    hallucination_response = "As an AI language model, I cannot provide that information."
+
+    print(f"Original: '{valid_response}'\nSanitized: '{sanitize(valid_response)}'")
+    print(f"\nOriginal: '{hallucination_response}'\nSanitized: '{sanitize(hallucination_response)}'")

rag/build_vector_store.py

@@ -1,6 +1,7 @@
 import sqlite3
 import chromadb
 from sentence_transformers import SentenceTransformer
+<<<<<<< HEAD
 from pathlib import Path
 import sys
 sys.path.append(str(Path(__file__).parent.parent))
@@ -10,10 +11,17 @@ def build_vector_store():
     """
     Builds a persistent vector store from the data in the SQLite database,
     embedding information about students, faculty, and courses.
+=======
+
+def build_vector_store():
+    """
+    Builds a persistent vector store from the data in the SQLite database.
+>>>>>>> feature/hf-deployment
     """
     conn = sqlite3.connect('database/university.db')
     cursor = conn.cursor()
 
+<<<<<<< HEAD
     documents = []
     print("Creating student docs")
     # === Build Student Documents ===
@@ -89,6 +97,35 @@ def build_vector_store():
     client = chromadb.PersistentClient(path="rag/vector_store")
     collection = client.get_or_create_collection("university_data")
 
+=======
+    # Fetch data from the database
+    students = cursor.execute("SELECT name, email FROM students").fetchall()
+    faculty = cursor.execute("SELECT name, email, department FROM faculty").fetchall()
+    courses = cursor.execute("SELECT name FROM courses").fetchall()
+
+    conn.close()
+
+    # Format data into documents
+    documents = []
+    for student in students:
+        documents.append(f"Student: {student[0]}, Email: {student[1]}")
+    for prof in faculty:
+        documents.append(f"Faculty: {prof[0]}, Email: {prof[1]}, Department: {prof[2]}")
+    for course in courses:
+        documents.append(f"Course: {course[0]}")
+
+    # Initialize the embedding model
+    model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
+
+    # Create embeddings
+    embeddings = model.encode(documents)
+
+    # Initialize ChromaDB client and create a collection
+    client = chromadb.PersistentClient(path="/tmp/vector_store")
+    collection = client.get_or_create_collection("university_data")
+
+    # Add documents and embeddings to the collection
+>>>>>>> feature/hf-deployment
     collection.add(
         embeddings=embeddings,
         documents=documents,
@@ -97,6 +134,9 @@ def build_vector_store():
 
     print("Vector store built successfully.")
 
+<<<<<<< HEAD
 
+=======
+>>>>>>> feature/hf-deployment
 if __name__ == "__main__":
     build_vector_store()

rag/retriever.py

@@ -1,9 +1,14 @@
 import chromadb
 from sentence_transformers import SentenceTransformer
 import sqlite3
+<<<<<<< HEAD
 from helper import get_similarity_model, sanitize
 
 def search(query: str, top_k: int = 10):
+=======
+
+def search(query: str, top_k: int = 5):
+>>>>>>> feature/hf-deployment
     """
     Searches the vector store for the most relevant documents to a given query.
     """
@@ -15,6 +20,7 @@ def search(query: str, top_k: int = 10):
         conn.close()
         return [{"name": student[0]} for student in students]
 
+<<<<<<< HEAD
     # get SentenceTransformer model
     model = get_similarity_model()
 
@@ -23,6 +29,16 @@ def search(query: str, top_k: int = 10):
 
     # Initialize ChromaDB client and get the collection
     client = chromadb.PersistentClient(path="rag/vector_store")
+=======
+    # Initialize the embedding model
+    model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
+
+    # Create the query embedding
+    query_embedding = model.encode([query])
+
+    # Initialize ChromaDB client and get the collection
+    client = chromadb.PersistentClient(path="/tmp/vector_store")
+>>>>>>> feature/hf-deployment
     collection = client.get_collection("university_data")
 
     # Perform the search
@@ -30,9 +46,14 @@ def search(query: str, top_k: int = 10):
         query_embeddings=query_embedding,
         n_results=top_k
     )
+<<<<<<< HEAD
     sanitized_context = [sanitize(doc) for doc in results['documents'][0]]
     return sanitized_context
 
+=======
+
+    return results['documents'][0]
+>>>>>>> feature/hf-deployment
 
 if __name__ == '__main__':
     # Example usage

requirements.txt

@@ -1,12 +1,11 @@
 streamlit==1.37.0
 python-dotenv==1.0.1
 transformers
-sentence-transformers
 scikit-learn
-faker
+Faker==15.3.4
 sentence-transformers==5.1.0
 numpy==1.26.4
 huggingface-hub==0.34.4
-chromadb
+chromadb==1.0.21
 langdetect
-nltk
+nltk

startup.sh

@@ -0,0 +1,4 @@
+#!/bin/bash
+
+# Run the Streamlit app
+streamlit run app.py