Journal Data Analysis and Visualization with LLMs

This project demonstrates how to analyze and visualize daily journal data using embeddings, dimensionality reduction, clustering, and an LLM for querying insights. It transforms structured journal entries into actionable narratives.

Features

• Load and preprocess journal data from a CSV file.
• Generate vector embeddings using OllamaEmbeddings.
• Apply PCA for dimensionality reduction and visualize clusters in 3D.
• Perform K-Means clustering to identify patterns in journal entries.
• Store embeddings in a vector database with Chroma.
• Query insights interactively using a locally hosted LLM.

Usage

• Prepare Journal Data: Save your daily logs in a CSV file (e.g., daily_track_records_2024.csv).
• Run Notebook: Follow the structured steps in the notebook:
• Load data and preprocess entries.
• Generate embeddings and perform clustering.
• Visualize journal data in 3D.
• Interact with the journal via LLM queries.

Query Example:

query = "What new skills or habits did I develop this year?"
response = chat_with_journal(query, vectorstore, llm)
print("LLM Response:", response)

Visualization

• Explore 3D scatter plots of PCA-reduced embeddings with clusters.
• Hover over points for detailed task information.

Acknowledgments

Inspired by daily journaling practices and leveraging the power of AI for personal reflection.