Skip to main content


This notebook covers how to load documents from Psychic. See here for more details.


  1. Follow the Quick Start section in this document
  2. Log into the Psychic dashboard and get your secret key
  3. Install the frontend react library into your web app and have a user authenticate a connection. The connection will be created using the connection id that you specify.

Loading documents

Use the PsychicLoader class to load in documents from a connection. Each connection has a connector id (corresponding to the SaaS app that was connected) and a connection id (which you passed in to the frontend library).

# Uncomment this to install psychicapi if you don't already have it installed
!poetry run pip -q install psychicapi langchain-chroma

[notice] A new release of pip is available: 23.0.1 -> 23.1.2
[notice] To update, run: pip install --upgrade pip
from langchain_community.document_loaders import PsychicLoader
from psychicapi import ConnectorId

# Create a document loader for google drive. We can also load from other connectors by setting the connector_id to the appropriate value e.g. ConnectorId.notion.value
# This loader uses our test credentials
google_drive_loader = PsychicLoader(

documents = google_drive_loader.load()

Converting the docs to embeddings

We can now convert these documents into embeddings and store them in a vector database like Chroma

from langchain.chains import RetrievalQAWithSourcesChain
from langchain_chroma import Chroma
from langchain_openai import OpenAI, OpenAIEmbeddings
from langchain_text_splitters import CharacterTextSplitter
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
texts = text_splitter.split_documents(documents)

embeddings = OpenAIEmbeddings()
docsearch = Chroma.from_documents(texts, embeddings)
chain = RetrievalQAWithSourcesChain.from_chain_type(
OpenAI(temperature=0), chain_type="stuff", retriever=docsearch.as_retriever()
chain({"question": "what is psychic?"}, return_only_outputs=True)

Help us out by providing feedback on this documentation page: