We see demand for tools that bridge the gap between prototyping and production. With usage based pricing and support for unlimited scaling, Pinecone Serverless helps to address pain points with vectorstore productionization that we've seen from the community. This repo builds a RAG chain that connects to Pinecone Serverless index using LCEL, turns it into an a web service with LangServe, uses Hosted LangServe deploy it, and uses LangSmith to monitor the input / outputs.
Follow instructions from Pinecone on setting up your serverless index.
Ensure these are set:
- PINECONE_API_KEY
- PINECONE_ENVIRONMENT
- PINECONE_INDEX_NAME
- OPENAI_API_KEY
Note: the choice of embedding model may require additional API keys, such as:
- COHERE_API_KEY
For prototyping:
poetry run jupyter notebook
This repo was created by following these steps:
(1) Create a LangChain app.
Run:
langchain app new .
This creates two folders:
app: This is where LangServe code will live
packages: This is where your chains or agents will live
It also creates:
Dockerfile: App configurations
pyproject.toml: Project configurations
Add your app dependencies to pyproject.toml and poetry.lock to support Pinecone serverless:
poetry add pinecone-client==3.0.0.dev8
poetry add langchain-community==0.0.12
poetry add cohere
poetry add openai
poetry add jupyter
Update enviorment based on the updated lock file:
poetry install
(2) Add your runnable (RAG app)
Create a file, chain.py with a runnable named chain that you want to execute.
This is our RAG logic (e.g., that we prototyped in our notebook).
Add chain.py to app directory.
Import the LCEL object in server.py:
from app.chain import chain as pinecone_wiki_chain
add_routes(app, pinecone_wiki_chain, path="/pinecone-wikipedia")
Run locally
poetry run langchain serve
(3) Deploy it with hosted LangServe
Go to your LangSmith console.
Select New Deployment.
Specify this Github url.
Add the abovementioned API keys as secrets.