Sending Embeddings

Create your own custom Item, Query, and User embeddings for use in all ranking models

Suggest Edits

Promoted supports sending your own embeddings on any entity. Set any attribute as a list of more than 30 floats or as a sufficiently long base64 encoded string of a float32 binary array. Use the same format as OpenAI embeddings. Promoted will automatically process these variables as embeddings. That's it!

Two examples of embeddings in a CMS or Properties JSON. "embedding1" uses the list of floats format. "embedding2" uses the base64 format.

{
  "embedding1": [-0.07205976,0.02388788,0.004621267,-0.01052015,-0.026126998,-0.06665136,-0.039351437,0.058605336,-0.15356456,-0.008707888,0.019232113,-0.02143768,0.001830364,-0.10805041,0.0053381897,0.00024992073,-0.010333571,-0.0462837,0.024370393,-0.04155039,-0.12155947,-0.06253068,-0.02006447,-0.0406716,-0.05272139,0.036312915,0.036812894,0.00092058233,-0.03117594,-0.09964825,-0.008561521,0.026015041,0.019604813,0.061004,-0.053690676,0.04826258,0.023055611,-0.02517642,-0.028567636,-0.04671914,0.064495124,-0.049844474,-0.04680418,0.040109176,0.07179293,-0.0110113025,-0.031239076,-0.030326363,-0.0015121714,0.012153916,-0.061114237,-0.022794714,-0.015041929,0.08881811,-0.048695207,0.046414904,0.066490516,-0.030903388,-0.012235642,-0.04845154,0.030548425,0.0014914124,-0.059624422,-0.07642649,-0.019556826,0.06671026,0.083138704,0.0069450843,0.0102588255,-0.06067494,-0.029734552,0.03652836,-0.15610075,0.059803385,-0.010370683,0.023005754,-0.09500467,0.015989803,0.0024970293,-0.044608787,0.033441916,0.05375884,-0.05246963,0.08011427,-0.025992986,0.011085631,-0.03451421,0.093687035,0.024262188,0.013318964,0.008646321,-0.04154606,0.08142801,-0.032021,0.056613278,0.045287944,-0.0075963205,-0.0788001,0.014076243,0.009144254,0.02583429,0.015052471,-0.03106613,0.001331797,0.007930704,0.02305757,0.05052777,0.025976619,-0.036625735,-0.035013765,0.014641192,0.022143694,0.003880236,-0.07074257,0.04371478,0.049295094,-0.09768892,0.06525048,-0.060744707,-0.03819729,-0.009088904,0.009527232,0.06290847,0.009384039,0.086939454,-0.081981435,-0.08989337,-1.1855612e-34,-0.05624016,0.06956535,0.020011058,0.005447706,0.00065015414,0.016140427,0.04882153,0.046335578,-0.075470224,-0.0014546785,0.01445186,-0.0077325692,-0.007343655,0.009517539,0.049272146,-0.03297804,0.026577126,0.061295435,-0.018501585,-0.048201956,0.040631935,-0.06259104,0.012280058,-0.037368737,0.09946315,0.067396276,-0.086162776,-0.00518845,0.107971646,0.0010528571,-0.02812807,-0.0044123936,-0.0070392033,0.03155531,0.02903939,-0.0124056395,-0.041254092,-0.11885312,0.0016916337,0.0057152533,-0.044593595,0.058644738,0.013138111,-0.044680033,-0.079754695,0.084490195,-0.009739361,0.036798548,0.02913927,0.071494006,-0.06334078,0.07253312,-0.043749087,0.007758503,-0.03218178,0.014491946,0.06908436,0.041797493,0.03937318,0.1356804,0.028214486,0.07887132,-0.038705382,0.08182371,-0.08241344,-0.01274963,0.014691253,-0.07816195,0.076736085,0.05095843,-0.031832717,-0.0057247896,-0.10027815,-0.05778764,0.031114962,-0.023160344,0.016842924,0.021115035,0.008328799,-0.059837744,-0.035042528,0.0100930035,-0.019268626,-0.07682053,0.0457127,-0.049767893,-0.011911384,-0.06671668,-0.022267835,-0.019079361,-0.013841473,0.0017243199,-0.05082028,0.033356853,-0.036661606,1.5940301e-34,-0.008253182,-0.02835566,0.03737778,0.11403897,0.07005319,4.1643892e-05,0.017963765,0.005893688,0.072961,0.083342366,-0.061149843,0.030975882,0.03162046,-0.06839984,0.027550386,-0.025141822,0.03213407,-0.09959931,0.01331843,0.03598377,-0.057843186,0.1001422,-0.012351363,0.032462455,-0.012342154,0.08381351,0.0137555525,0.08127492,-0.0389969,-0.010873791,-0.07162653,0.028638896,0.023588771,-0.08112644,-0.023042643,0.010487639,0.1177842,0.09582711,-0.04011568,0.016385186,0.073182516,0.014004782,-0.040751766,0.045171864,-0.10115692,0.008995974,-0.120089635,0.10984123,-0.029911276,0.02791277,-0.0028953368,0.072076246,0.015096972,0.04483014,0.0015998172,0.014829258,0.08949116,-0.008129441,-0.065542914,0.042949293,-0.018956076,-0.016182436,0.063306816,0.027002992,0.0786024,-0.08421495,0.04949523,-0.06460858,-0.063930444,0.023083143,0.05266332,-0.0028763073,0.044541843,-0.041455135,0.00010465424,-0.041471567,-0.055604987,-0.066254765,0.02093254,0.065724626,-0.028101273,-0.042142846,0.053653,0.06196477,-0.027504569,0.03737002,0.075896084,-0.00408588,-0.0196223,0.009535103,-0.07296515,-0.02496726,-0.035596993,0.10154543,-0.040130306,-1.9606343e-08,-0.021799108,-0.03112594,-0.026642436,0.07939269,0.07234696,-0.026922783,-0.090515435,0.10299534,-0.028810114,0.035253927,-0.0033626913,-0.061053526,-0.025573205,0.0025459677,0.103784546,0.010860129,-0.014790837,0.009059937,-0.11588336,-0.03488949,0.05363463,0.05752889,0.03043691,-0.0017236206,0.06089027,0.012384883,0.0425415,0.011583149,0.005324918,-0.007329942,0.0014599205,-0.022803625,0.023918439,0.08442021,-0.002013742,-0.014154661,0.035276804,0.043196123,0.031092169,-0.014220844,-0.019134713,0.018683081,-0.08189191,0.0036358817,-0.08554946,-0.009830383,-0.041473724,-0.032265678,-0.0032702903,-0.047731567,-0.013130204,-0.03210582,0.0061169122,0.0355428,0.070048444,0.064011715,0.04127366,-0.016045988,0.047316723,0.08427212,0.049523905,-0.09386123,0.057357017,0.05891776],
	"embedding2": "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"
}

How Promoted Uses Embeddings

In Delivery, all pairs embeddings with the same number of dimensions generate several derivative features automatically:

  • Dot product
  • Distribution features for that dot product for all similarly computed dot products in that Request. Distribution features include percentile, normalized value, mean, max, and median.
  • The L2 norm

This elegantly produces Item X User and Item X Query embeddings of many different types of media without configuration. Promoted may also include the embeddings themselves into models as features when our AutoML detects that this improves model fit.

What Embeddings Represent

Promoted is agnostic about what the embeddings represent. They could have been generated by text, images, or structured data like JSON. There can be any number of embeddings, and they can appear in any hierarchical structure in a JSON document (e.g., a nested map or ordered list are acceptable formats).

Where to Send Embeddings

Anywhere a feature is set can set an embedding.

  • Content CMS: Any Item. If you set the contentID to a query string, you can also set your own query embeddings.
  • User CMS: Any User embeddings.
  • Delivery API: Embeddings can be sent on Request.Properties or Request.Insertion.Properties

Embedding Versions

Generally, retraining embeddings causes incompatibility with past versions of embeddings when using embeddings as features in models directly and when computing dot products. If you have important embeddings and want to change the embedding model, send both the new and old versions of the embedding as different key names. Then, inform Promoted about the version migration. We will help you safely transition from the old to new version of embedding. Once the old version is deprecated, you can stop sending it to Promoted to save resources.

Updated about 2 months ago