This project is a lightweight AI-powered assistant designed to enhance the learning experience of students in higher education. Users can ask questions in natural language based on uploaded study materials (PDFs), and receive intelligent, context-aware answers directly sourced from the selected topic.
Built with LangChain, Chroma, and OpenAI, the system leverages large language models and a retrieval-augmented generation (RAG) architecture to provide accurate, conversational answers through a minimal Streamlit interface.
The assistant supports:
- Ingesting and indexing lecture PDFs using Chroma for efficient semantic search.
- Real-time Q&A powered by OpenAI’s GPT model.
- Streamlit UI for easy interaction.
Ideal for test prep, revision, or flipped classroom models.
- Boost student engagement and comprehension.
- Provide an always-available academic Q&A tool.
- Demonstrate RAG with LangChain + Chroma + OpenAI.
- Quick development using modern LLM tooling.
- Runs locally—just Python, internet, and a laptop.
- Ideal for 2-week hackathons, demos, or prototypes.
- Admin topic creation and PDF upload
- Student topic selection and Q&A
- SQLite database for managing users and chat sessions
- Continued use of LangChain, Chroma, and OpenAI
| Feature | Responsibility | Implemented By |
|---|
| PDF Parsing | Break PDF into clean text chunks | load_and_split_pdf()
|
| Vector Store | Embed & store by topic ID | create_chroma_index_for_topic()
|
| LLM Chain | Ask topic-specific questions | build_qa_chain()
|
| Admin Portal | Create topic, upload study material | admin_topic_upload_ui()
|
| Student UI | Select topic, ask questions | student_qa_ui()
|
| SQLite DB | Manage users, topics, sessions, messages | setup_sqlite_database()
|
document_loader.py
from langchain.document_loaders import PyPDFLoader
from langchain.text_splitter import CharacterTextSplitter
def load_and_split_pdf(file_path, chunk_size=500, chunk_overlap=50):
loader = PyPDFLoader(file_path)
pages = loader.load()
splitter = CharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
return splitter.split_documents(pages)
vector_store.py
from langchain.vectorstores import Chroma
from langchain.embeddings.openai import OpenAIEmbeddings
PERSIST_DIR = "chroma_db"
def create_chroma_index(chunks, persist_directory=PERSIST_DIR):
embeddings = OpenAIEmbeddings()
vectorstore = Chroma.from_documents(documents=chunks, embedding=embeddings, persist_directory=persist_directory)
vectorstore.persist()
def load_chroma_retriever(persist_directory=PERSIST_DIR):
embeddings = OpenAIEmbeddings()
vectorstore = Chroma(persist_directory=persist_directory, embedding_function=embeddings)
return vectorstore.as_retriever()
qa_chain.py
from langchain.chains import RetrievalQA
from langchain.chat_models import ChatOpenAI
def build_qa_chain(retriever):
llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
return RetrievalQA.from_chain_type(llm=llm, retriever=retriever)
- Always assume the dev server is running and avoid starting the server