We're excited to announce the release of Aquila HUB 0.2.0.
This release achieves 10x performance upgrades to support your traffic needs.
We're continuing and constantly evaluating experimental IPFS support in Aquila HUB. Now you can run an IPFS node and leverage the decentralized distribution of your Fasttext models. We are also planning to add support to more language models in future releases.
A first look at the Aquila HUB IPFS connection
To ensure full control over IPFS nodes and their management, we have decoupled Aquila HUB from a running IPFS node. Aquila HUB uses an IPFS gateway connection to communicate with your local or any pre-configured public node. Learn more about IPFS gateway.
Configuring Aquila HUB to use your IPFS gateway
Make sure that you have cloned the latest release of Aquila HUB from Github. Modify
AquilaHub/src/config.yml file with the correct gateway URL.
Loading a model from IPFS to generate embeddings
Loading an IPFS model is very similar to specifying an HTTP endpoint for your Fasttext model in schema definition.
Make sure to follow this URL format:
ftxt:ipfs://<IPFS content ID of a fasttext .bin model file> to download files from IPFS network.
Aquila HUB will take care of downloading the model and then loading it into the memory later on.