Thought Embedding Application

This repository contains a simple Streamlit application for embedding and analyzing user thoughts using PostgreSQL and Principal Component Analysis (PCA).

Table of Contents

• Cloud Setting
• • API
• • Queries
• Usage
• Functions Overview
  • import_embedding.py
  • stream_app.py
  • utils.py
• License

Cloud Setting

1. Create a Google Cloud Account
2. Create an AlloyDB Cluster
3. Create an AlloyDB instance

API

Enable below APIs:

1. AlloyDB API
2. Compute Engine API
3. Cloud Resource Manager API
4. Service Networking API

Queries

Create table

CREATE TABLE "public".thought_embedding ( thought TEXT, thought_id SERIAL PRIMARY KEY, entry_date TIMESTAMP );

Enable below extensions

CREATE EXTENSION IF NOT EXISTS vector CREATE EXTENSION IF NOT EXISTS google_ml_integration

Query to autogenerate vectors

ALTER TABLE thought_embedding ADD COLUMN embedding vector GENERATED ALWAYS AS (embedding('textembedding-gecko@001',thought)) STORED;

Functions Overview

import_embedding.py

• getconn(): Establishes connection to PostgreSQL using Google Cloud AlloyDB.

• inserttodb(thought: str): Inserts a user thought into the PostgreSQL database.

• getembedding(): Retrieves all stored thought embeddings from the database.

• similar_thoughts(thought): Finds top 3 similar thoughts based on vector similarity.

stream_app.py

• User Interface: Collects and records thoughts, displaying similar entries and PCA plot.

utils.py

• compute_pca(): Performs PCA on embeddings, returning principal components for visualization.

Requirements

• Python 3.8+
• PostgreSQL
• Google Cloud AlloyDB
• Required Python packages:

pip install sqlalchemy streamlit pandas matplotlib scikit-learn pg8000