Personal Journal RAG

This is a repository for a Personal Journal RAG (Retrieval-Augmented Generation) system. It uses a personal journal as a knowledge base and leverages a language model to generate responses based on the content of the journal.

Usage

You will need to install the required dependencies by running the following command:

pip install -r requirements.txt

You will then need to transform your journal into CSV. My journal is a single txt file that has the following format:

Jan 1, 2025

10:00 - this is an entry
10:20 - this is an entry

Jan 2, 2025

10:00 - this is an entry
...

The journalToCsv.py file is adapted to this format. You can modify it to suit your needs.

python journalToCsv.py journal.txt

After you generate the CSV file, you can use the run generate_index.py to create a journal_index.faiss file.

python3 generate_index.py

Then, you can use the run search_journal.py to generate a response based on your journal. First, you need to have ollama running on localhost:11434. The default model is llama3:latest.

python3 search_journal.py "your question here"

Example:

python3 search_journal.py "What concert did I buy tickets for?"

The model will generate search terms, look for them in the text and come up with a final answer:

Final Answer:
Based on the provided journal entries, the name of the band you bought tickets for is Twenty One Pilots. The entries specifically mention that "There is a band called Balu Brigada pra will open the Twenty One Pilots concert" (Jan 22, 2025, Time: 16:57) and that "Nous avons acheté les tickets pour l'événement de demain. On y va à 14:30 et Gabe va avec nous" (Jan 20, 2024, Time: 20:36), which implies that you bought tickets for a Twenty One Pilots concert.