Skip to content

Python client for easy use of the Recombee recommendation API

Notifications You must be signed in to change notification settings

recombee/python-api-client

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

RecombeeApiClient

A Python 3 client for easy use of the Recombee recommendation API.

If you don't have an account at Recombee yet, you can create a free account here.

Documentation of the API can be found at docs.recombee.com.

Installation

Install the client with pip:

$ pip install recombee-api-client

Examples

Basic example

from recombee_api_client.api_client import RecombeeClient, Region
from recombee_api_client.exceptions import APIException
from recombee_api_client.api_requests import *
import random

client = RecombeeClient('--my-database-id--', '--db-private-token--', region=Region.US_WEST)

#Generate some random purchases of items by users
PROBABILITY_PURCHASED = 0.1
NUM = 100
purchase_requests = []

for user_id in ["user-%s" % i for i in range(NUM) ]:
  for item_id in ["item-%s" % i for i in range(NUM) ]:
    if random.random() < PROBABILITY_PURCHASED:

      request = AddPurchase(user_id, item_id, cascade_create=True)
      purchase_requests.append(request)

try:
    # Send the data to Recombee, use Batch for faster processing of larger data
    print('Send purchases')
    client.send(Batch(purchase_requests))

    # Get recommendations for user 'user-25'
    response = client.send(RecommendItemsToUser('user-25', 5))
    print("Recommended items: %s" % response)

    # User scrolled down - get next 3 recommended items
    response = client.send(RecommendNextItems(response['recommId'], 3))
    print("Next recommended items: %s" % response)

except APIException as e:
    print(e)

Using property values

from recombee_api_client.api_client import RecombeeClient, Region
from recombee_api_client.api_requests import AddItemProperty, SetItemValues, AddPurchase
from recombee_api_client.api_requests import RecommendItemsToItem, SearchItems, Batch, ResetDatabase
import random

NUM = 100
PROBABILITY_PURCHASED = 0.1

client = RecombeeClient('--my-database-id--', '--db-private-token--', region=Region.AP_SE)

# Clear the entire database
client.send(ResetDatabase())

# We will use computers as items in this example
# Computers have four properties
#   - price (floating point number)
#   - number of processor cores (integer number)
#   - description (string)
#   - image (url of computer's photo)

# Add properties of items
client.send(AddItemProperty('price', 'double'))
client.send(AddItemProperty('num-cores', 'int'))
client.send(AddItemProperty('description', 'string'))
client.send(AddItemProperty('image', 'image'))

# Prepare requests for setting a catalog of computers
requests = [SetItemValues(
    "computer-%s" % i, #itemId
    #values:
    {
      'price': random.uniform(500, 2000),
      'num-cores': random.randrange(1,9),
      'description': 'Great computer',
      'image': 'http://examplesite.com/products/computer-%s.jpg' % i
    },
    cascade_create=True   # Use cascadeCreate for creating item
                          # with given itemId if it doesn't exist
  ) for i in range(NUM)]


# Send catalog to the recommender system
client.send(Batch(requests))

# Prepare some purchases of items by users
requests = []
items = ["computer-%s" % i for i in range(NUM)]
users = ["user-%s" % i for i in range(NUM)]

for item_id in items:
    #Use cascadeCreate to create unexisting users
    purchasing_users = [user_id for user_id in users if random.random() < PROBABILITY_PURCHASED]
    requests += [AddPurchase(user_id, item_id, cascade_create=True) for user_id in purchasing_users]

# Send purchases to the recommender system
client.send(Batch(requests))

# Get 5 recommendations for user-42, who is currently viewing computer-6
# Recommend only computers that have at least 3 cores
recommended = client.send(
    RecommendItemsToItem('computer-6', 'user-42', 5, filter="'num-cores'>=3")
)
print("Recommended items with at least 3 processor cores: %s" % recommended)

# Recommend only items that are more expensive then currently viewed item (up-sell)
recommended = client.send(
    RecommendItemsToItem('computer-6', 'user-42', 5, filter="'price' > context_item[\"price\"]")
)
print("Recommended up-sell items: %s" % recommended)

# Filters, boosters and other settings can be also set in the Admin UI (admin.recombee.com)
# when scenario is specified
recommended = client.send(
  RecommendItemsToItem('computer-6', 'user-42', 5, scenario='product_detail')
  )

# Perform personalized full-text search with a user's search query (e.g. 'computers').
matches = client.send(SearchItems('user-42', 'computers', 5, scenario='search_top'))
print("Matched items: %s" % matches)

Exception handling

For the sake of brevity, the above examples omit exception handling. However, various exceptions can occur while processing request, for example because of adding an already existing item, submitting interaction of nonexistent user or because of timeout.

We are doing our best to provide the fastest and most reliable service, but production-level applications must implement a fallback solution since errors can always happen. The fallback might be, for example, showing the most popular items from the current category, or not displaying recommendations at all.

Example:

from recombee_api_client.exceptions import *

try:
  recommended = client.send(
      RecommendItemsToItem('computer-6', 'user-42', 5, filter="'price' > context_item[\"price\"]")
  )
except ResponseException as e:
  #Handle errorneous request => use fallback
except ApiTimeoutException as e:
  #Handle timeout => use fallback
except APIException as e:
  #APIException is parent of both ResponseException and ApiTimeoutException

About

Python client for easy use of the Recombee recommendation API

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages