Example where LangChain uses Ollama to summarize PDF documents

Here's a simple example where LangChain uses Ollama to summarize text from a PDF document locally.

Prerequisites

1. Install Ollama and download a model (like llama2 or mistral):

   # Install Ollama
   brew install ollama
   
   # Download Mistral model
   ollama pull mistral
   

2. Install required Python packages:

   pip install langchain pypdf
   

---

Code Example

from langchain.document_loaders import PyPDFLoader
from langchain.llms import Ollama
from langchain.chains.summarize import load_summarize_chain

# Step 1: Load the PDF document
pdf_path = "sample.pdf"  # Replace with your file path
loader = PyPDFLoader(pdf_path)
documents = loader.load()

# Step 2: Initialize the Ollama LLM
llm = Ollama(model="mistral")  # Use your downloaded model

# Step 3: Create the Summarization Chain
summarize_chain = load_summarize_chain(llm, chain_type="map_reduce")

# Step 4: Generate the Summary
summary = summarize_chain.run(documents)
print("Summary:")
print(summary)

---

How This Works

1. PyPDFLoader reads and splits the PDF into smaller text chunks.
2. The map_reduce chain first summarizes each chunk independently (map phase).
3. Then it combines those smaller summaries into one final summary (reduce phase).
4. LangChain sends each chunk to the Ollama API for text generation.

---

🔑 Advantages

• Local processing = 100% privacy.
• Works offline.
• No external API keys needed.
• Fast inference on consumer hardware.

---

Bonus Tip 💡

If you want streaming summaries (token-by-token output), modify the LLM call like this:

for chunk in llm.stream("Summarize the following text: " + documents[0].page_content):
    print(chunk, end="")