AI PDF Chatbot With LangChain 2026: Build One in 30 Minutes

No Docker, no heavy frameworks - here's a minimal LangChain PDF chatbot you can build in 30 minutes and extend into a real product.

💡 What You Will Learn

No Docker, no heavy frameworks - here's a minimal LangChain PDF chatbot you can build in 30 minutes and extend into a real product.

📜 Table of Contents

• The Minimal RAG Pattern
• The Stack
• The Core Code
• Why Each Piece Is There
• The 30-Minute Path
• FAQ

The Minimal RAG Pattern

A PDF chatbot is a RAG pipeline with four steps: load, split, embed, retrieve-and-answer. LangChain (144,172 stars, 2026-08-14) gives you each step as a component; the whole bot is about 60 lines.

The Stack

• Python 3.10+
• langchain + langchain-community + langchain-chroma
• pypdf (PDF parsing)
• A vector store: Chroma (local, free, zero-config)
• An LLM: any OpenAI-compatible endpoint - OpenAI, DeepSeek, or local Ollama

The Core Code

from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_chroma import Chroma
from langchain_openai import ChatOpenAI, OpenAIEmbeddings

loader = PyPDFLoader("manual.pdf")
docs = loader.load()
splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
chunks = splitter.split_documents(docs)
vectorstore = Chroma.from_documents(chunks, OpenAIEmbeddings(model="text-embedding-3-small"))
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
retriever = vectorstore.as_retriever(search_kwargs={"k": 4})

while True:
    q = input("Ask: ")
    if q.lower() in ("exit", "quit"):
        break
    docs = retriever.invoke(q)
    context = "
".join(d.page_content for d in docs)
    print(llm.invoke(f"Answer using ONLY the context. If unsure, say so.
Context: {context}
Question: {q}"))

Why Each Piece Is There

• Chunk size 1000 / overlap 200: answers need surrounding context; overlap prevents cutting a paragraph in half.
• k=4 retrieval: four chunks is enough for most answers without bloating the prompt.
• temperature=0: factual Q&A should be deterministic.
• The prompt constraint: 'only context, admit uncertainty' stops hallucinations from dressing up as answers.

The 30-Minute Path

1. pip install langchain langchain-community langchain-chroma pypdf langchain-openai (5 min)
2. Save the script, point PyPDFLoader at your PDF (2 min)
3. Run, ask questions (10 min of playing)
4. Wrap it in FastAPI (102,000 stars) and add a tiny chat UI (rest of the time)

FAQ

Why LangChain instead of building raw? Components handle the fiddly parts (splitting, retrieval, prompt assembly); raw code is educational, LangChain is productive.

Which embedding model? text-embedding-3-small is cheap and good enough; local users can use Ollama embeddings.

Does this work with scanned PDFs? No - PyPDFLoader needs text; add OCR (pytesseract) for scans.

How accurate is it? Depends on chunk quality and the LLM; the k=4 retrieval with source citations is the accuracy lever.