update readme with pinecone info

Author: wjayesh

llm-complete-guide/README.md

@@ -18,7 +18,7 @@ using ZenML, enabling you to build powerful, scalable, and maintainable
 LLM-powered applications.
 
 This project contains all the pipeline and step code necessary to follow along
-with the guide. You'll need a PostgreSQL database to store the embeddings; full
+with the guide. You'll need a vector store to store the embeddings; full
 instructions are provided below for how to set that up.
 
 ## 📽️ Watch the webinars
@@ -55,7 +55,40 @@ zenml secret create llm-complete --openai_api_key=<your-openai-api-key>
 export ZENML_PROJECT_SECRET_NAME=llm-complete
 ```
 
-### Setting up Supabase
+### Setting up Pinecone
+
+[Pinecone](https://www.pinecone.io/) is the default vector store used in this project. It's a cloud-native vector database that's optimized for machine learning applications. You'll need to create a Pinecone account and get an API key to use it.
+
+Once you have your Pinecone account set up, you'll need to store your API key and index name as a ZenML secret (with name `pinecone-zenml`). You can do this by running the following command:
+
+```shell
+zenml secret create pinecone-zenml --pinecone_api_key=<YOUR_PINECONE_API_KEY> --pinecone_env=<YOUR_PINECONE_ENV> --pinecone_index=<YOUR_INDEX_NAME>
+```
+
+The `pinecone_index` value you specify will be used for all your development pipeline runs. When you promote your ZenML model to production and run your ingestion pipeline again, it will automatically create a new production index called `<YOUR_INDEX_NAME>-prod`. This separation ensures that your development and production environments remain isolated.
+
+### Choosing Your Vector Store
+
+While Pinecone is the default vector store, this project supports multiple vector stores. You can choose between:
+
+1. **Pinecone** (default): A cloud-native vector database optimized for machine learning applications
+2. **PostgreSQL with pgvector**: A local or cloud PostgreSQL database with vector similarity search capabilities
+3. **Elasticsearch**: A distributed search engine with vector search support
+
+To switch between vector stores, you need to create or modify a pipeline configuration file (e.g., `configs/dev/rag.yaml`) and set the `index_type` parameter for the `index_generator` step. For example:
+
+```yaml
+steps:
+  index_generator:
+    parameters:
+        index_type: pinecone  # Options: pinecone, postgres, elasticsearch
+```
+
+This configuration will be used by both the basic RAG and RAG pipelines. Each vector store requires its own setup and credentials as described in their respective sections below.
+
+### Alternative: Setting up Supabase
+
+While Pinecone is the default vector store, you can still use Supabase's PostgreSQL database as an alternative. 
 
 [Supabase](https://supabase.com/) is a cloud provider that offers a PostgreSQL
 database. It's simple to use and has a free tier that should be sufficient for
@@ -76,7 +109,7 @@ string from the Supabase dashboard.
 
 ![](.assets/supabase-connection-string.png)
 
-In case Supabase is not an option for you, you can use a different database as the backend.
+In case neither Pinecone nor Supabase is an option for you, you can use a different database as the backend.
 
 ### Running the RAG pipeline
 
@@ -89,12 +122,12 @@ python run.py rag
 ```
 
 This will run the basic RAG pipeline, which scrapes the ZenML documentation and
-stores the embeddings in the Supabase database.
+stores the embeddings in your configured vector store (Pinecone by default).
 
 ### Querying your RAG pipeline assets
 
-Once the pipeline has run successfully, you can query the assets in the Supabase
-database using the `--query` flag as well as passing in the model you'd like to
+Once the pipeline has run successfully, you can query the assets in your vector store
+using the `--query` flag as well as passing in the model you'd like to
 use for the LLM.
 
 When you're ready to make the query, run the following command: