Dynamic Embeddings

Let’s look at an example of how this could be used:

from incribo import DynamicEmbeddingWrapper

# Create or add your existing embeddings
emb = DynamicEmbeddingWrapper([0.4140090346336365, 0.7410697340965271,  
0.9770076274871826, 0.04037611931562424, 0.1512177288532257])

# Get the current vector
current_vector = emb.get_vector()
print(current_vector)  # Output: [0.414009034, 0.7410...]

# Update the embedding
emb.update([4.0, 5.0, 6.0])

# Get the updated vector
updated_vector = emb.get_vector()
print(updated_vector)  # Output: [4.0, 5.0, 6.0]

In the example above, we perform function calls of creating a new embedding using DynamicEmbeddingWrapper which stores the vector(or “state”) in memory.

When the underlying data changes, we can auto-update the embeddings using the update() method to change the state of your vectors.

NOTE: This method does not delete or erase your previous embeddings(more on state fallbacks in VersionControlWrapper).

This component does not perform any computations on the embeddings itself - instead, it acts as a container that facilitates the storage and updating of embedding vectors.
The actual computation of new embeddings would typically be done externally, and then the results can be stored in this wrapper.