An experimental Retrieval Augmented Generation (RAG) AI on Cloudflare Data Localization Suite (DLS).
- Cloudflare Wrangler
- Cloudflare Workers AI
- Cloudflare D1
- Cloudflare Vectorize
- LangChain JS
The entire knowledge base data on Cloudflare Data Localization Suite (DLS) is stored in the Cloudflare D1 database dls_data across multiple tables.
The Model / Embeddings API used for the knowledge base data is @cf/baai/bge-base-en-v1.5 with 768 output dimensions using distance metrics cosine.
Cosine similarity: Cosine similarity is a measure of the similarity between two vectors. It is calculated by taking the dot product of the two vectors and dividing by the product of their norms. Cosine similarity is a good choice for semantic text retrieval because it is insensitive to the order of the terms in the vectors. This means that it can be used to compare documents that are written in different ways, but that have the same meaning.
Source: Vector DB Architecture for a custom documents chatbot
Visit dls-ai.dlsdemo.com/vector-store to retrieve the configuration of the Vectorize Vector Store, including its configured dimensions and distance metric.
The Generative Text Model / Embedding Model used is @cf/meta/llama-2-7b-chat-fp16.
Review available Embedding Models.
Visit dls-ai.dlsdemo.com and add your question to the ?query= parameter:
https://dls-ai.dlsdemo.com/?query=Which%20regions%20does%20Regional%20Services%20support%3F
Recommended use is through the UI ai.dlsdemo.com.
By running the /demo endpoint, the Workers queries a D1 database to data to the Vectorize Vector Store: our Knowledge Base.
In order to clear / delete the entire Knowledge Base (Vectorize Vector Store), simply run the /clear endpoint.
Retrieve the Vector Store details:
/vector-store
Run a Vector Search similaritySearchWithScore for similar documents in the knowledge base:
/vector-search?query=what%20is%20regional%20services?
Get streaming responses:
/streaming?query=what%20is%20regional%20services?
When inside the working directory, install all dependencies:
npm install
Run a local server for developing:
npx wrangler dev --remote
Deploy the project:
npx wrangler deploy
Keep the package and models up2date by periodically reviewing the DevDocs and running:
npm update @cloudflare/ai --save-dev
For educational purposes only.
This is an experimental project using Retrieval Augmented Generation (RAG) AI on Cloudflare Workers AI. It is not intended for production use. Use at your own risk. Beware of hallucinations.
The answers provided by this RAG AI are often incorrect, incomplete, or lacking in context.