Unli.ai docs

Appearance

Pinecone ​

Adding a Vector Store to a Project ​

Each project in Unli.ai can have at most one vector store. Follow these steps to add a Pinecone vector store to your project:

  1. Navigate to the Project section.
  2. Select Vector Store from the project settings.
  3. Choose Pinecone as your vector store provider.
  4. Enter your Pinecone API Key.
    • If you are using a custom Pinecone instance, enter the custom host as well.
  5. Save the configuration.

Once configured, your vector store will be ready for use.

Data Ingestion & Automatic Storage ​

Once data is ingested into Unli.ai through any of the following methods:

  • Datasource Integrations
  • Direct Upload
  • Web Scraping

The system automatically adds the data to the vector store.

Automated Processing with Beesight ​

Beesight handles:

  • Pagination
  • Background processing
  • Data indexing

However, you can override these defaults by providing custom parameters:

json
{
    "datasource": "ds_01JME43RDQEMN801ADPDXWJEK2",
    "documents": [
        {
            "id": "12Im3HMIahS3Iq8hnJ40ZHM-U4N-6fWvl"
        },
        {
            "id": "1eg3CaU9ytf5QH38FuaxAyQgmFN2HYZFbBN7E3l9kZZg"
        }
    ],
    "chunking_params": {
        "chunk_size": 3333,
        "chunk_overlap": 23
    }
}

Document Metadata in Vector Store ​

When documents are added to the vector store, the following metadata is automatically stored:

  • document_url
  • document_name
  • document_id
  • document_type
  • document_source (Datasource ID from Unli.ai)
  • internal_id (Traceback ID in our system)
  • document_body (Text data of the document)
  • document_page_number
  • document_total_pages
  • chunk_index

Custom Metadata ​

You can also add custom metadata by passing additional key-value pairs in the API payload while adding documents.