Mycelia is a powerful Similarity Search Engine that helps developers quickly build and integrate AI-based recommendations, content search and much more.
Get up and runing with your similarity queries with just a few lines of code - in any language that make API calls!
# <<< PYTHON >>> import requests ## Full notebook: https://www.kaggle.com/dionisio/self- supervised-beer-similarity-api/ # Let's query the most similar beer, # from the Kaggle Craft Beers Dataset db_name = 'beers2021' query_id = 1439 top_k = 2 # number one is always the query itself # ID 1439 # Brunette Nut Brown Ale # English Brown Ale # ABV: 0.048 # IBU 15.0 # query mycelia API mycelia_url = f'https://mycelia.azure-api.net /clone/similar/id/{db_name}?id={query}&top_k={top_k}' # GET answers ans = requests.get(mycelia_url) # ans: { # "similarity": [{ # "query_id": 1439, # "results": [{ # "id": 1439, # "distance": 0.0 # }, { # "id": 2148, # "distance": 12.52 # }] # }] # } # ID 2148 # Saddle Bronc Brown Ale (2013) # English Brown Ale # ABV: 0.048 # IBU 16.0
Mycelia works with *any* kind of data: tables, documents (text) and images. Here's a similarity query with some beers to make your day go smoother.
The community edition is intended for evaluational purposes and can be accessed directly from the Mycelia API page.
No. As Mycelia is a REST API, you could use any language that you love and unleash the full power of the Mycelia Engine.
It means that it's not a cool AI proof-of-concept, but a low-latency engine that runs on top of reliable and secure infrastructure, designed to handle bilions of data entries and thousands of requests per second.
Mycelia was designed to efficiently extract semantic information from data and serve it as a low-latency API.
The main use cases range from content-based or collaborative-filtering recommendation engines - recommending similar items to similar customers, to image search on websites/apps and Entity Matching - when there is no shared key between tables and you need to find the best match between entities.
Mycelia is backed by modern deep learning algorithms that process data and turn them into meaninful latent vectors, which then can be queried very efficiently by means of ANN - Approximate Nearest Neighbor.