# LangGraph Agent with Memory and Tool Confirmation

This project demonstrates a simple conversational agent built with **LangGraph** that:

- **Remembers** the conversation history across turns using an in‑memory checkpoint.
- **Pauses** before invoking any tool, asking the user for confirmation.
- Uses the **Rich** library for pretty console output.

## Prerequisites

- Python 3.10+
- An OpenAI API key (set as the `OPENAI_API_KEY` environment variable).

## Installation

```bash
# Clone the repository (or copy the files)
git clone https://github.com/yourusername/langgraph-agent.git
cd langgraph-agent

# Create a virtual environment (optional but recommended)
python -m venv .venv
source .venv/bin/activate   # On Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt
```

## Usage

```bash
python src/main.py
```

You will see a prompt:

```
Start chat. Type 'exit' to quit.
You:
```

Type any message. If the agent decides to use a tool (e.g., the `echo` tool), it will pause and ask:

```
Agent wants to call tool: echo({"text":"Hello"})
Разрешить? (Y/n):
```

- Type `Y` or press Enter to allow the tool to run.
- Type `n` to cancel the tool call.

The conversation history is preserved across turns, so the agent can refer back to earlier messages.

## Customizing

- **Tools**: Add more tools by defining functions decorated with `@tool` from `langchain.tools`.
- **Memory**: Replace `MemorySaver()` with a persistent checkpoint (e.g., `RedisSaver`) for long‑term storage.
- **Prompt**: Modify the `system_prompt` in `create_agent` to change the agent’s behavior.

## License

MIT License