This application creates a Retrieval-Augmented Generation (RAG) system that allows users to ask questions about their custom-made diary. For demonstration, the system uses an arbitrarily created diary, with AI generated entries stored in a .csv file. The script for injecting this data appropriately into a PostgreSQL database with vector embeddings is provided, it enables context-aware responses to user queries.

• Python 3.8+
• PostgreSQL 12+ with pgvector extension
• OpenAI API key

1. Clone this repository and change directory to its root folder:

git clone https://github.com/JoeCardoso13/diary_bot.git && cd diary_bot

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate

3. Install dependencies:

pip install -r requirements.txt

4. Set up PostgreSQL:

   • Install PostgreSQL if you haven't already
   • Install the pgvector extension:

     CREATE EXTENSION vector;

   • Create a database named diary_bot

5. Add your OpenAI API key into your environment, e.g. an .env file with the following:

OPENAI_API_KEY=your_api_key_here
DB_NAME=diary_bot
DB_USER=your_psql_user_name
DB_HOST=/var/run/postgresql_or_localhost
DB_PORT=5432

Before using the system, you need to process the diary entries by running the pre_processing.py script:

python src/pre_processing.py

This will:

• Read the CSV file
• Generate embeddings for each entry
• Store the data in the PostgreSQL database

Run the Gradio interface for a user-friendly web experience:

python src/main_gradio_interface.py

This will:

• Start a local web server
• Open the interface in your default browser
• Allow you to:
  • Ask questions about my hike
  • Get random question suggestions
  • See responses with source attribution

Run the CLI version for a simpler interface:

python src/main_command_line.py

This provides:

• A text-based interface
• Same functionality as the web version
• Responses in the terminal

diary_bot/
├── data/
│ └── diary.csv # Blog entries data
├── src/
│ ├── pre_processing.py # Data processing and embedding generation
│ ├── rag_tools.py # Shared RAG functionality
│ ├── main_command_line.py # CLI interface
│ └── main_gradio_interface.py # Web interface
├── requirements.txt # Python dependencies
└── README.md

The diary table contains:

• id: Unique identifier
• date: Date of diary entry
• entry: Content of diary entry
• embedding: Vector embedding of the entry text

The next evolution of Diary Bot leans into the Model Context Protocol.

The idea: expose the existing RAG system as an MCP server, then surface it as an MCP App — an iframe that lives inside a claude.ai session. The agent there can see everything the app renders, query the diary freely, and reason over it in real time. The RAG system stops being a standalone tool and becomes something a live Claude agent can reach into on demand.

The gut feeling driving this: a diary is personal context. An agent with access to your personal context — able to surface patterns, make connections, answer questions you didn't know to ask — is a fundamentally different kind of assistant.