Retrieval Augmented Generation

Overview

This is an example node.js application that utilizes local embeddings and Ollama for text retrieval and response generation. It processes a text corpus, generates embeddings for "chunks" using transformers.js, and uses these embeddings to perform a "similarity search" in response to queries.

• server.js: Server file that handles API requests and integrates with Ollama for LLM inference.
• save-embeddings.js: Process a text file and generate embeddings using transformers.js.
• embeddings.json: Precomputed embeddings generated from the text corpus.
• public/: p5.js sketch
• .env: Configuration file (no API tokens required)

References

• Transformers.js Documentation
• Ollama Documentation

Instructions

1. Install and run Ollama

ollama run gemma3

2. Install Dependencies

npm install

3. Generate embeddings: Run save-embeddings.js to create the embeddings.json file.

The code reads from data/processing-course.txt by default. To use your own text corpus, modify the filename and adjust how the text is split:

const raw = fs.readFileSync('data/your-text-file.txt', 'utf-8');
let chunks = raw.split(/\n+/); // Adjust splitting logic as needed

Then run:

node save-embeddings.js

4. Run the server:

node server.js

Open your browser to: http://localhost:3001 (or whatever port is specified)

• All processing runs locally - no external API keys required
• First run will download the transformers.js models (cached for future use)
• The embedding model is lightweight (~50MB)
• You can use any model available in Ollama for LLM inference