Web Research Assistant

A terminal-based AI research assistant built with Python, LangChain, OpenRouter, and Rich.

It can:

• Summarize webpages into study notes
• Chat with your saved notes via session memory
• Run entirely from the terminal

---

Features

Web Summarization

Load any webpage:

/url https://docs.langchain.com/

The assistant:

1. Downloads the webpage
2. Extracts readable text
3. Generates concise study notes
4. Loads the notes into the current session

---

Chat With Notes

After loading a webpage:

> what is langchain?

> explain tools

> compare chains and agents

The assistant answers using the summarized notes as context.

---

Session Memory

Memory stores only the current session (URL, summary, and message history).
Starting a new URL overwrites the previous session. Exiting wipes all memory.

---

Exit Behavior

When exiting:

/exit

Memory is wiped immediately and the assistant exits.

---

Installation

Clone

git clone <your-repo-url>
cd web-loader

Install Dependencies

Using UV:

uv sync

Or manually:

uv add \
    langchain \
    langchain-openrouter \
    beautifulsoup4 \
    requests \
    python-dotenv \
    rich

---

Configuration

Create your environment file:

cp .env.example .env

Add your OpenRouter API key:

MISTRAL_API_KEY=your_api_key_here

---

Running

Start the assistant:

uv run main.py

You should see:

Web Research Assistant

Type /help

---

Commands

Summarize a webpage

/url https://overthewire.org/wargames/bandit/

Help

/help

Exit (wipes memory)

/exit

---

Project Structure

web-loader/
│
├── main.py
├── chatbot.py
├── chat.py
├── summarizer.py
├── memory.py
├── config.py
├── cli.py
│
├── data/
│   └── memory.json
│
├── .env.example
├── README.md
├── pyproject.toml
└── .gitignore

---

Example Workflow

> /url https://overthewire.org/wargames/bandit/

(summary displayed in green panel)

> what does bandit teach?

(Answer displayed in cyan panel)

> /exit

Goodbye!

---

Future Improvements

• PDF support
• YouTube transcript summarization
• Multi-document chat
• Vector database retrieval
• Local embeddings
• Streaming responses
• Export notes to Markdown
• Export notes to PDF
• Conversation history
• Multi-model support
• Source citations