Insert Documents

APIs

To insert a new document, you first need to create a new collection.

Once you have a collection, you can start inserting documents using the following API:

curl -X PATCH \
  http://localhost:8080/v0/collections/{COLLECTION_ID}/documents \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer <write_api_key>' \
  -d '{
    "title": "My first document",
    "content": "This is the content of my first document."
  }'

OramaCore will automatically generate generate text embeddings for the document in real-time, allowing you to perform full-text, hybrid, and vector search immediately.

Optimizations

After inserting a document, OramaCore will perform a number of optimizations.

Since OramaCore uses two immutable data structures (FST for full-text search and HNSW for vector search), inserting a new document won't automatically guarantee that it will be stored in such structures immediately.

OramaCore will insert all the new documents into a buffer, and once the buffer reaches a certain size, it will be flushed to the immutable data structures.

The buffer size is configurable via the insert_batch_commit_size configuration in the config.yaml file.

By default, OramaCore will commit (so will push the buffer to the immutable data structures) every 5000 documents. You can change this value to better suit your use case.

One thing to note is that OramaCore will be able to search for the document immediately after inserting it, even if it's not yet in the immutable data structures. This is because OramaCore will also store the document in a mutable data structure, allowing you to search for it immediately.

And since the buffer size is tipically small (we recommend < 100k elements), you won't notice any performance degradation.