Ollama phi3 based text summarizer

🧠 Build a Text Summarizer Using Local Ollama (phi3) + LangChain

In this tutorial, we’ll build a text summarizer using:

• 🦙 Ollama (running locally)
• 🤖 phi3 model
• 🔗 LangChain
• 🐍 Python

By the end, you’ll have a working CLI-based text summarizer that runs fully offline.

---

📌 What We’re Building

Input: Long text

Output: Clean summarized text

Model: phi3 running locally via Ollama

---

🖥️ Step 1: Install Ollama

Linux / macOS

curl -fsSL https://ollama.com/install.sh | sh

Windows

Download installer from:
https://ollama.com

---

🚀 Step 2: Pull the phi3 Model

After installation:

ollama pull phi3

Test it:

ollama run phi3

Type something like:

Summarize: Artificial Intelligence is transforming the world...

If it responds → ✅ Ollama is working.

Exit with:

/exit

---

🐍 Step 3: Create Python Environment

mkdir text_summarizer_project
cd text_summarizer_project

python -m venv venv
source venv/bin/activate  # Linux/macOS
# OR
venv\Scripts\activate     # Windows

---

📦 Step 4: Install Required Packages

pip install langchain langchain-ollama langchain-core

---

📁 Step 5: Project Structure

Create this structure:

text_summarizer_project/
│
├── main.py
└── utils/
    ├── loader.py
    ├── write_output.py
    └── constants.py

---

🧱 Step 6: Create Supporting Files

---

📝 utils/constants.py

information = """
Artificial Intelligence (AI) is a branch of computer science
that aims to create systems capable of performing tasks
that normally require human intelligence...
"""

summary_template = """
You are a helpful AI assistant.

Summarize the following text in a concise paragraph:

{information}
"""

---

💾 utils/write_output.py

def write_output(text, filename):
    with open(filename, "w", encoding="utf-8") as f:
        f.write(text)

---

⏳ utils/loader.py (Simple Terminal Spinner)

import sys
import threading
import time

class Loader:
    def __init__(self, message, success, fail, color="cyan"):
        self.message = message
        self.success = success
        self.fail = fail
        self.done = False

    def spinner(self):
        while not self.done:
            for char in "|/-\\":
                sys.stdout.write(f"\r{self.message} {char}")
                sys.stdout.flush()
                time.sleep(0.1)

    def __enter__(self):
        self.thread = threading.Thread(target=self.spinner)
        self.thread.start()

    def __exit__(self, exc_type, exc_value, traceback):
        self.done = True
        self.thread.join()
        if exc_type:
            print(f"\r❌ {self.fail}")
        else:
            print(f"\r✅ {self.success}")

---

🧠 Step 7: Create main.py

from langchain_core.prompts import PromptTemplate
from langchain_ollama import ChatOllama
from utils.loader import Loader
from utils.write_output import write_output
from utils.constants import information, summary_template


def text_summarizer():
    print("🧠 Local Text Summarizer Using phi3")

    summary_prompt_template = PromptTemplate(
        input_variables=["information"],
        template=summary_template
    )

    llm = ChatOllama(
        model="phi3",
        temperature=0
    )

    chain = summary_prompt_template | llm

    try:
        with Loader(
            message="Agent is thinking...",
            success="Response received!",
            fail="Model crashed!"
        ):
            response = chain.invoke({"information": information})

        write_output(response.content, "output.txt")
        print("📄 Summary written to output.txt")

    except Exception as e:
        print("Error:", e)


if __name__ == "__main__":
    text_summarizer()

---

▶️ Step 8: Run the Project

Make sure:

• Ollama is installed
• phi3 is pulled
• Your virtual environment is activated

Run:

python main.py

You should see:

Agent is thinking...
✅ Response received!
Summary written to output.txt

Open output.txt to see the result.

---

🧪 How It Works

1️⃣ PromptTemplate

Formats the user input into a proper instruction.

2️⃣ ChatOllama

Connects to locally running Ollama model.

3️⃣ LCEL Operator (|)

This line:

chain = summary_prompt_template | llm

Means:

Prompt → LLM → Response

---

🔥 Optional Improvements

You can extend this project by:

• Accepting user input dynamically
• Adding a CLI argument parser
• Building a FastAPI wrapper
• Converting into a web app
• Streaming responses instead of waiting

---

⚠️ Troubleshooting

❌ ModuleNotFoundError

Check:

• You activated virtual environment
• All files exist
• You are running from correct directory

❌ Model not found

Run:

ollama pull phi3

---

🎯 Final Result

You now have a fully offline:

• LLM-powered
• Terminal-based
• Extensible text summarizer

Running locally on your machine 🔥