This connector extracts documents from an appropriately configured Notion site using the Notion API and the LlamaIndex loader for Notion. It relies on a configured Azure OpenAI embedding model to get the appropriate vector embeddings for input documents.
A Notion integration's API token is required for this connector. Follow these instructions to create a Notion integration and retrieve the associated API secret.
The integration will only have access to pages that it has been added to, or pages whose parent it has been added to. To ensure a certain section or group of pages is crawled, add the integration to the parent of those pages by following these instructions.
Additionally, this connector requires Azure OpenAI services or an OpenAI API key to generate embedding vectors for documents.
Create a YAML config file based on the following template.
notion_api_token: <notion_api_token>
embedding_model:
azure_openai:
key: <key>
endpoint: <endpoint>Note that an embedding model needs to be appropriately configured. This example shows how to configure an Azure OpenAI services model, but you can use other supported models.
These defaults are provided; you don't have to manually configure them.
include_text refers to if you'd like to include the original document text alongside the embedded content.
embedding_model: # in the same block as above
azure_openai:
version: <version> # "2024-03-01-preview"
deployment_name: <deployment_name> # "Embedding_3_small"
model: <model> # "text-embedding-3-small"
chunk_size: 512
chunk_overlap: 50
notion_api_version: <api_key_version> # "2022-06-28"
include_text: <include_text> # FalseSee Output Config for more information on the optional output config.
Follow the installation instructions to install metaphor-connectors in your environment (or virtualenv). Make sure to include the notion or all extra.
Run the following command to test the connector locally:
metaphor notion <config_file>Manually verify the output after the run finishes.