From a77d21d7aa9db3fe31151a8491b86c7260f16d02 Mon Sep 17 00:00:00 2001
From: "mich.kolomanski" <mich.kolomanski@gmail.com>
Date: Thu, 24 Mar 2022 09:17:34 +0100
Subject: [PATCH] [#144] Integrate ElasticSearch

---
 CHANGELOG.md                                  |   1 +
 Pipfile                                       |   2 +-
 Pipfile.lock                                  | 335 +++++++++---------
 recommender/api/schemas/recommendation.py     |   9 +
 .../engines/base/base_inference_component.py  |  41 ++-
 .../ncf/inference/ncf_inference_component.py  |  11 +-
 .../rl/inference/rl_inference_component.py    |  13 +-
 .../engines/rl/ml_components/sars_encoder.py  |   2 +-
 .../rl/ml_components/sarses_generator.py      |   4 +
 .../rl/ml_components/search_data_encoder.py   |  57 ---
 .../rl/ml_components/service_encoder.py       |  84 +++++
 .../engines/rl/ml_components/state_encoder.py |  40 ++-
 .../synthetic_dataset/dataset.py              | 139 +-------
 .../model_evaluation_step.py                  |   3 +-
 recommender/engines/rl/utils.py               |   8 +-
 recommender/errors.py                         |   4 +-
 recommender/models/recommendation.py          |   1 +
 recommender/models/state.py                   |   1 +
 recommender/services/deserializer.py          |   1 +
 recommender/services/fts.py                   |  28 +-
 recommender/utils.py                          |  56 +--
 tests/endpoints/test_recommendations.py       |   1 +
 .../preprocessing/test_search_data_encoder.py | 117 ------
 .../preprocessing/test_service_encoder.py     | 132 +++++++
 .../ml_components/test_normalizer.py          |   1 -
 .../inference/test_ncf_inference_component.py |  35 +-
 .../inference/test_rl_inference_component.py  |  48 ++-
 .../synthetic_dataset/test_dataset.py         |   7 -
 tests/services/test_fts.py                    |  57 +--
 29 files changed, 601 insertions(+), 637 deletions(-)
 delete mode 100644 recommender/engines/rl/ml_components/search_data_encoder.py
 create mode 100644 recommender/engines/rl/ml_components/service_encoder.py
 delete mode 100644 tests/engine/preprocessing/test_search_data_encoder.py
 create mode 100644 tests/engine/preprocessing/test_service_encoder.py

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 1ffba179..3a8d9a2e 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased] - YYYY-MM-DD
 ### Added
+- Integrate elastic_search results from the Marketplace [@Michal-Kolomanski]
 - JMS Subscriber for user actions and flask cli for it [@michal-szostak]
 - search_data migration [@Michal-Kolomanski]
 - tqdm and logging in embedding pipeline [@Michal-Kolomanski]
diff --git a/Pipfile b/Pipfile
index 504aea0d..09c4d4ae 100644
--- a/Pipfile
+++ b/Pipfile
@@ -15,7 +15,6 @@ markupsafe = "==2.0.1"
 flask-restx = "==0.3.0"
 jsonref = "*"
 json-ref-dict = "*"
-pymongo = "*"
 mongoengine = "*"
 flask-mongoengine = "*"
 inflection = "*"
@@ -38,6 +37,7 @@ sentry-sdk = {extras = ["flask"], version = "*"}
 ray = {version = "==1.7.0", extras = ["default"]}
 celery = {version = "==5.1.2", extras = ["redis"]}
 "stomp.py" = "==7.0.0"
+pymongo = "==3.12"
 pydantic = "*"
 
 [dev-packages]
diff --git a/Pipfile.lock b/Pipfile.lock
index 504ccb0e..d08added 100644
--- a/Pipfile.lock
+++ b/Pipfile.lock
@@ -1,7 +1,7 @@
 {
     "_meta": {
         "hash": {
-            "sha256": "64129a64ea77e53f567a10da3bdc45b93a5e49b7b8793a08fb3a28e5c0e3f3b9"
+            "sha256": "33b131d8d785100c0c34d51fdf628948a5c1a41124158eec65db8fade280f225"
         },
         "pipfile-spec": 6,
         "requires": {
@@ -147,14 +147,6 @@
             "markers": "python_full_version >= '3.6.2'",
             "version": "==3.5.0"
         },
-        "appnope": {
-            "hashes": [
-                "sha256:02bd91c4de869fbb1e1c50aafc4098827a7a54ab2f39d9dcba6c9547ed920e24",
-                "sha256:265a455292d0bd8a72453494fa24df5a11eb18373a60c7c0430889f22548605e"
-            ],
-            "markers": "platform_system == 'Darwin'",
-            "version": "==0.1.3"
-        },
         "argon2-cffi": {
             "hashes": [
                 "sha256:8c976986f2c5c0e5000919e6de187906cfd81fb1c72bf9d88c01177e77da7f80",
@@ -249,7 +241,7 @@
                 "sha256:08a1fe86d253b5c88c92cc3d810fd8048a16d15762e1e5b74d502256e5926aa1",
                 "sha256:c6d6cc054bdc9c83b48b8083e236e5f00f238428666d2ce2e083eaa5fd568565"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==5.0.0"
         },
         "blessings": {
@@ -378,7 +370,7 @@
                 "sha256:a0713dc7a1de3f06bc0df5a9567ad19ead2d3d5689b434768a6145bff77c0667",
                 "sha256:f184f0d851d96b6d29297354ed981b7dd71df7ff500d82fa6d11f0856bee8035"
             ],
-            "markers": "python_version < '4.0' and python_full_version >= '3.6.2'",
+            "markers": "python_full_version >= '3.6.2' and python_full_version < '4.0.0'",
             "version": "==0.3.0"
         },
         "click-plugins": {
@@ -494,7 +486,7 @@
                 "sha256:e3513399177dd37af4c1332df52da5da1d0c387e5927dc4c0709e26ee7302e8f",
                 "sha256:eb1946efac0c0c3d411cea0b5ac772fbde744109fd9520fb0c5a51979faf05ad"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==1.6.0"
         },
         "decorator": {
@@ -526,7 +518,7 @@
                 "sha256:0f7569a4a6ff151958b64304071d370daa3243d15941a7beedf0c9fe5105603e",
                 "sha256:a851e51367fb93e9e1361732c1d60dab63eff98712e503ea7d92e6eccb109b4f"
             ],
-            "markers": "python_version >= '3.6' and python_version < '4.0'",
+            "markers": "python_version >= '3.6' and python_full_version < '4.0.0'",
             "version": "==2.2.1"
         },
         "docopt": {
@@ -586,7 +578,7 @@
                 "sha256:9cd540a9352e432c7246a48fe4e8712b10acb1df2ad1f30e8c070b82ae1fed85",
                 "sha256:f8314284bfffbdcfa0ff3d7992b023d4c628ced6feb957351d4c48d059f56bc0"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==3.6.0"
         },
         "flask": {
@@ -623,11 +615,11 @@
         },
         "fonttools": {
             "hashes": [
-                "sha256:14b94aab22439b9bd46c8f44b057b1d2f4b7ebe86556f9b15a1b25694bb12519",
-                "sha256:65d14ab1fe70cbd2f18fca538d98bd45d73e9b065defb843da71dc3c454deb45"
+                "sha256:50f0c76985321412123a1c32d1017ccad0e4ab1c71c8e64797cd7ffcbb7a379f",
+                "sha256:66dd757b8a9d7b07d61c22b8d973282990c45343572b88e5fc722b3a8272598e"
             ],
-            "markers": "python_full_version >= '3.7.0'",
-            "version": "==4.33.0"
+            "markers": "python_version >= '3.7'",
+            "version": "==4.33.1"
         },
         "frozenlist": {
             "hashes": [
@@ -691,7 +683,7 @@
                 "sha256:f96293d6f982c58ebebb428c50163d010c2f05de0cde99fd681bfdc18d4b2dc2",
                 "sha256:ff9310f05b9d9c5c4dd472983dc956901ee6cb2c3ec1ab116ecdde25f3ce4951"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==1.3.0"
         },
         "google-api-core": {
@@ -903,7 +895,7 @@
                 "sha256:0e28273e290858393e86e152b104e5506a79c13d25b951ac6eca220051b4be60",
                 "sha256:2b0987af43c0d4b62cecb13c592755f599f96f29aafe36c01731aaa96df30d39"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==6.13.0"
         },
         "ipython": {
@@ -911,7 +903,7 @@
                 "sha256:468abefc45c15419e3c8e8c0a6a5c115b2127bafa34d7c641b1d443658793909",
                 "sha256:86df2cf291c6c70b5be6a7b608650420e89180c8ec74f376a34e2dc15c3400e7"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==7.32.0"
         },
         "ipython-genutils": {
@@ -989,7 +981,7 @@
                 "sha256:636694eb41b3535ed608fe04129f26542b59ed99808b4f688aa32dcf55317a83",
                 "sha256:77281a1f71684953ee8b3d488371b162419767973789272434bbc3f29d9c8823"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==4.4.0"
         },
         "jupyter-client": {
@@ -997,7 +989,7 @@
                 "sha256:44045448eadc12493d819d965eb1dc9d10d1927698adbb9b14eb9a3a4a45ba53",
                 "sha256:8fdbad344a8baa6a413d86d25bbf87ce21cb2b4aa5a8e0413863b9754eb8eb8a"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==7.2.2"
         },
         "jupyter-core": {
@@ -1005,7 +997,7 @@
                 "sha256:a6de44b16b7b31d7271130c71a6792c4040f077011961138afed5e5e73181aec",
                 "sha256:e7f5212177af7ab34179690140f188aa9bf3d322d8155ed972cbded19f55b6f3"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==4.10.0"
         },
         "jupyter-server": {
@@ -1013,7 +1005,7 @@
                 "sha256:72dd1ff5373d2def94e80632ba4397e504cc9200c5b5f44b5b0af2e062a73353",
                 "sha256:c756f87ad64b84e2aa522ef482445e1a93f7fe4a5fc78358f4636e53c9a0463a"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==1.16.0"
         },
         "jupyterlab": {
@@ -1029,7 +1021,7 @@
                 "sha256:2405800db07c9f770863bcf8049a529c3dd4d3e28536638bd7c1c01d2748309f",
                 "sha256:7405d7fde60819d905a9fa8ce89e4cd830e318cdad22a0030f7a901da705585d"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.2.2"
         },
         "jupyterlab-server": {
@@ -1086,7 +1078,7 @@
                 "sha256:fa4d97d7d2b2c082e67907c0b8d9f31b85aa5d3ba0d33096b7116f03f8061261",
                 "sha256:ffbdb9a96c536f0405895b5e21ee39ec579cb0ed97bdbd169ae2b55f41d73219"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==1.4.2"
         },
         "kombu": {
@@ -1094,7 +1086,7 @@
                 "sha256:37cee3ee725f94ea8bb173eaab7c1760203ea53bbebae226328600f9d2799610",
                 "sha256:8b213b24293d3417bcf0d2f5537b7f756079e3ea232a8386dcc89a59fd2361a4"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==5.2.4"
         },
         "markdown": {
@@ -1345,7 +1337,7 @@
                 "sha256:feba80698173761cddd814fa22e88b0661e98cb810f9f986c54aa34d281e4937",
                 "sha256:feea820722e69451743a3d56ad74948b68bf456984d63c1a92e8347b7b88452d"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==6.0.2"
         },
         "nbclassic": {
@@ -1353,7 +1345,7 @@
                 "sha256:36dbaa88ffaf5dc05d149deb97504b86ba648f4a80a60b8a58ac94acab2daeb5",
                 "sha256:89184baa2d66b8ac3c8d3df57cbcf16f34148954d410a2fb3e897d7c18f2479d"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.3.7"
         },
         "nbclient": {
@@ -1361,7 +1353,7 @@
                 "sha256:40c52c9b5e3c31faecaee69f202b3f53e38d7c1c563de0fadde9d7eda0fdafe8",
                 "sha256:47ac905af59379913c1f8f541098d2550153cf8dc58553cbe18c702b181518b0"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.5.13"
         },
         "nbconvert": {
@@ -1369,7 +1361,7 @@
                 "sha256:21163a8e2073c07109ca8f398836e45efdba2aacea68d6f75a8a545fef070d4e",
                 "sha256:e01d219f55cc79f9701c834d605e8aa3acf35725345d3942e3983937f368ce14"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==6.4.5"
         },
         "nbformat": {
@@ -1377,7 +1369,7 @@
                 "sha256:38856d97de49e8292e2d5d8f595e9d26f02abfd87e075d450af4511870b40538",
                 "sha256:fcc5ab8cb74e20b19570b5be809e2dba9b82836fd2761a89066ad43394ba29f5"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==5.3.0"
         },
         "nest-asyncio": {
@@ -1393,7 +1385,7 @@
                 "sha256:709b1856a564fe53054796c80e17a67262071c86bfbdfa6b96aaa346113c555a",
                 "sha256:b4a6baf2eba21ce67a0ca11a793d1781b06b8078f34d06c710742e55f3eee505"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==6.4.11"
         },
         "notebook-shim": {
@@ -1401,7 +1393,7 @@
                 "sha256:02432d55a01139ac16e2100888aa2b56c614720cec73a27e71f40a5387e45324",
                 "sha256:7897e47a36d92248925a2143e3596f19c60597708f7bef50d81fcd31d7263e85"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.1.0"
         },
         "numpy": {
@@ -1580,7 +1572,7 @@
                 "sha256:f401ed2bbb155e1ade150ccc63db1a4f6c1909d3d378f7d1235a44e90d75fb97",
                 "sha256:fb89397013cf302f282f0fc998bb7abf11d49dcff72c8ecb320f76ea6e2c5717"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==9.1.0"
         },
         "pluggy": {
@@ -1632,7 +1624,7 @@
                 "sha256:e250a42f15bf9d5b09fe1b293bdba2801cd520a9f5ea2d7fb7536d4441811d20",
                 "sha256:ff8d8fa42675249bb456f5db06c00de6c2f4c27a065955917b28c4f15978b9c3"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==3.20.1"
         },
         "psutil": {
@@ -1796,92 +1788,107 @@
         },
         "pymongo": {
             "hashes": [
-                "sha256:019a4c13ef1d9accd08de70247068671b116a0383adcd684f6365219f29f41cd",
-                "sha256:07f50a3b8a3afb086089abcd9ab562fb2a27b63fd7017ca13dfe7b663c8f3762",
-                "sha256:08a619c92769bd7346434dfc331a3aa8dc63bee80ed0be250bb0e878c69a6f3e",
-                "sha256:0a3474e6a0df0077a44573727341df6627042df5ca61ea5373c157bb6512ccc7",
-                "sha256:0b8a1c766de29173ddbd316dbd75a97b19a4cf9ac45a39ad4f53426e5df1483b",
-                "sha256:0f7e3872fb7b61ec574b7e04302ea03928b670df583f8691cb1df6e54cd42b19",
-                "sha256:17df40753085ccba38a0e150001f757910d66440d9b5deced30ed4cc8b45b6f3",
-                "sha256:298908478d07871dbe17e9ccd37a10a27ad3f37cc1faaf0cc4d205da3c3e8539",
-                "sha256:302ac0f4825501ab0900b8f1a2bb2dc7d28f69c7f15fbc799fb26f9b9ebb1ecb",
-                "sha256:303d1b3da2461586379d98b344b529598c8156857285ba5bd156dab1c875d1f6",
-                "sha256:306336dab4537b2343e52ec34017c3051c3aee5a961fff4915ab27f7e6d9b1e9",
-                "sha256:30d35a8855f328a85e5002f0908b24e500efdf8f5f78b73098995ce111baa2a9",
-                "sha256:3139c9ddee379c22a9109a0b3bf4cdb64597db2bbd3909f7a2825b47226977a4",
-                "sha256:32e785c37f6a0e844788c6085ea2c9c0c528348c22cebe91896705a92f2b1b26",
-                "sha256:33a5693e8d1fbb7743b7e867d43c1095652a0c6fedddab6cefe6020bee2ca393",
-                "sha256:35d02603c2318676fca5049cdc722bb2e7a378eaccf139ad767365e0eb3bcdbe",
-                "sha256:4516a5ce2beaebddc74d6e304ed520324dda99573c310ef4078284b026f81e93",
-                "sha256:49bb36986f11da2da190a2e777a411c0a28eeb8623850091ea8099b84e3860c7",
-                "sha256:4aa4800530782f7d38aeb169476a5bc692aacc394686f0ca3866e4bb85c9aa3f",
-                "sha256:4d1cdece06156542c18b691511a01fe78a694b9fa287ffd8e15680dbf2beeed5",
-                "sha256:4e4d2babb8737d650250d0fa940ffa1b88aa92b8eb399af093734950a1eeca45",
-                "sha256:4fd5c4f25d8d488ee5701c3ec786f52907dca653b47ce8709bcc2bfb0f5506ae",
-                "sha256:52c8b7bffd2140818ade2aa28c24cfe47935a7273a3bb976d1d8fb17e716536f",
-                "sha256:56b856a459762a3c052987e28ed2bd4b874f0be6671d2cc4f74c4891f47f997a",
-                "sha256:571a3e1ef4abeb4ac719ac381f5aada664627b4ee048d9995e93b4bcd0f70601",
-                "sha256:5cae9c935cdc53e4729920543b7d990615a115d85f32144773bc4b2b05144628",
-                "sha256:5d6ef3fa41f3e3be93483a77f81dea8c7ce5ed4411382a31af2b09b9ec5d9585",
-                "sha256:6396f0db060db9d8751167ea08f3a77a41a71cd39236fade4409394e57b377e8",
-                "sha256:69beffb048de19f7c18617b90e38cbddfac20077b1826c27c3fe2e3ef8ac5a43",
-                "sha256:7507439cd799295893b5602f438f8b6a0f483efb00720df1aa33a39102b41bcf",
-                "sha256:7aa40509dd9f75c256f0a7533d5e2ccef711dbbf0d91c13ac937d21d76d71656",
-                "sha256:7d69a3d980ecbf7238ab37b9027c87ad3b278bb3742a150fc33b5a8a9d990431",
-                "sha256:7dae2cf84a09329617b08731b95ad1fc98d50a9b40c2007e351438bd119a2f7a",
-                "sha256:7f36eacc70849d40ce86c85042ecfcbeab810691b1a3b08062ede32a2d6521ac",
-                "sha256:7f55a602d55e8f0feafde533c69dfd29bf0e54645ab0996b605613cda6894a85",
-                "sha256:8357aa727094798f1d831339ecfd8b3e388c01db6015a3cbd51790cb75e39994",
-                "sha256:84dc6bfeaeba98fe93fc837b12f9af4842694cdbde18083f150e80aec3de88f9",
-                "sha256:86b18420f00d5977bda477369ac85e04185ef94046a04ae0d85f5a807d1a8eb4",
-                "sha256:89f32d8450e15b0c11efdc81e2704d68c502c889d48415a50add9fa031144f75",
-                "sha256:8a1de8931cdad8cd12724e12a6167eef8cb478cc3ee5d2c9f4670c934f2975e1",
-                "sha256:8f106468062ac7ff03e3522a66cb7b36c662326d8eb7af1be0f30563740ff002",
-                "sha256:9a4ea87a0401c06b687db29e2ae836b2b58480ab118cb6eea8ac2ef45a4345f8",
-                "sha256:9ee1b019a4640bf39c0705ab65e934cfe6b89f1a8dc26f389fae3d7c62358d6f",
-                "sha256:a0d7c6d6fbca62508ea525abd869fca78ecf68cd3bcf6ae67ec478aa37cf39c0",
-                "sha256:a1417cb339a367a5dfd0e50193a1c0e87e31325547a0e7624ee4ff414c0b53b3",
-                "sha256:a35f1937b0560587d478fd2259a6d4f66cf511c9d28e90b52b183745eaa77d95",
-                "sha256:a4a35e83abfdac7095430e1c1476e0871e4b234e936f4a7a7631531b09a4f198",
-                "sha256:a7d1c8830a7bc10420ceb60a256d25ab5b032a6dad12a46af6ab2e470cee9124",
-                "sha256:a938d4d5b530f8ea988afb80817209eabc150c53b8c7af79d40080313a35e470",
-                "sha256:a9a2c377106fe01a57bad0f703653de286d56ee5285ed36c6953535cfa11f928",
-                "sha256:baf7546afd27be4f96f23307d7c295497fb512875167743b14a7457b95761294",
-                "sha256:bb21e2f35d6f09aa4a6df0c716f41e036cfcf05a98323b50294f93085ad775e9",
-                "sha256:bc62ba37bcb42e4146b853940b65a2de31c2962d2b6da9bc3ce28270d13b5c4e",
-                "sha256:be3ba736aabf856195199208ed37459408c932940cbccd2dc9f6ff2e800b0261",
-                "sha256:c03eb43d15c8af58159e7561076634d565530aaacaf48cf4e070c3501e88a372",
-                "sha256:c1349331fa743eed4042f9652200e60596f8beb957554acbcbb42aad4272c606",
-                "sha256:c3637cfce519560e2a2579d05eb81e912d109283b8ddc8de46f57ec20d273d92",
-                "sha256:c481cd1af2a77f58f495f7f87c2d715c6f1179d07c1ec927cca1f7977a2d99aa",
-                "sha256:c575f9499e5f540e034ff87bef894f031ae613a98b0d1d3afcc1f482527d5f1c",
-                "sha256:c604831daf2e7e5979ecd97a90cb8c4a7bae208ff45bc792e32eae09c3281afb",
-                "sha256:c759e1e0333664831d8d1d6b26cf59f23f3707758f696c71f506504b33130f81",
-                "sha256:c8a2743dd50629c0222f26c5f55975e45841d985b4b1c7a54b3f03b53de3427d",
-                "sha256:cbcac9263f500da94405cc9fc7e7a42a3ba6c2fe88b2cd7039737cba44c66889",
-                "sha256:cce1b7a680653e31ff2b252f19a39f1ded578a35a96c419ddb9632c62d2af7d8",
-                "sha256:cf96799b3e5e2e2f6dbca015f72b28e7ae415ce8147472f89a3704a035d6336d",
-                "sha256:d06ed18917dbc7a938c4231cbbec52a7e474be270b2ef9208abb4d5a34f5ceb9",
-                "sha256:d4ba5b4f1a0334dbe673f767f28775744e793fcb9ea57a1d72bc622c9f90e6b4",
-                "sha256:d7b8f25c9b0043cbaf77b8b895814e33e7a3c807a097377c07e1bd49946030d5",
-                "sha256:d86511ef8217822fb8716460aaa1ece31fe9e8a48900e541cb35acb7c35e9e2e",
-                "sha256:db8a9cbe965c7343feab2e2bf9a3771f303f8a7ca401dececb6ef28e06b3b18c",
-                "sha256:dbe92a8808cefb284e235b8f82933d7d2e24ff929fe5d53f1fd3ca55fced4b58",
-                "sha256:deb83cc9f639045e2febcc8d4306d4b83893af8d895f2ed70aa342a3430b534c",
-                "sha256:df9084e06efb3d59608a6a443faa9861828585579f0ae8e95f5a4dab70f1a00f",
-                "sha256:dfb89e92746e4a1e0d091cba73d6cc1e16b4094ebdbb14c2e96a80320feb1ad7",
-                "sha256:e13ddfe2ead9540e8773cae098f54c5206d6fcef64846a3e5042db47fc3a41ed",
-                "sha256:e4956384340eec7b526149ac126c8aa11d32441cb3ce77a690cb4821d0d0635c",
-                "sha256:e6eecd027b6ba5617ea6af3e12e20d578d8f4ad1bf51a9abe69c6fd4835ea532",
-                "sha256:eff9818b7671a55f1ce781398607e0d8c304cd430c0581fbe15b868a7a371c27",
-                "sha256:f0aea377b9dfc166c8fa05bb158c30ee3d53d73f0ed2fc05ba6c638d9563422f",
-                "sha256:f1fba193ab2f25849e24caa4570611aa2f80bc1c1ba791851523734b4ed69e43",
-                "sha256:f6db4f00d3baad615e99a865539391243d12b113fb628ebda1d7794ce02d5a10",
-                "sha256:f9405c02af86850e0a8a8ba777b7e7609e0d07bff46adc4f78892cc2d5456018",
-                "sha256:fb4445e3721720c5ca14c0650f35c263b3430e6e16df9d2504618df914b3fb99"
+                "sha256:02dc0b0f48ed3cd06c13b7e31b066bf91e00dac5f8147b0a0a45f9009bfab857",
+                "sha256:053b4ebf91c7395d1fcd2ce6a9edff0024575b7b2de6781554a4114448a8adc9",
+                "sha256:070a4ef689c9438a999ec3830e69b208ff0d12251846e064d947f97d819d1d05",
+                "sha256:072ba7cb65c8aa4d5c5659bf6722ee85781c9d7816dc00679b8b6f3dff1ddafc",
+                "sha256:0b6055e0ef451ff73c93d0348d122a0750dddf323b9361de5835dac2f6cf7fc1",
+                "sha256:11f9e0cfc84ade088a38df2708d0b958bb76360181df1b2e1e1a41beaa57952b",
+                "sha256:18290649759f9db660972442aa606f845c368db9b08c4c73770f6da14113569b",
+                "sha256:186104a94d39b8412f8e3de385acd990a628346a4402d4f3a288a82b8660bd22",
+                "sha256:1970cfe2aec1bf74b40cf30c130ad10cd968941694630386db33e1d044c22a2e",
+                "sha256:19d4bd0fc29aa405bb1781456c9cfff9fceabb68543741eb17234952dbc2bbb0",
+                "sha256:1bab889ae7640eba739f67fcbf8eff252dddc60d4495e6ddd3a87cd9a95fdb52",
+                "sha256:1bc6fe7279ff40c6818db002bf5284aa03ec181ea1b1ceaeee33c289d412afa7",
+                "sha256:208debdcf76ed39ebf24f38509f50dc1c100e31e8653817fedb8e1f867850a13",
+                "sha256:2399a85b54f68008e483b2871f4a458b4c980469c7fe921595ede073e4844f1e",
+                "sha256:246ec420e4c8744fceb4e259f906211b9c198e1f345e6158dcd7cbad3737e11e",
+                "sha256:24f8aeec4d6b894a6128844e50ff423dd02462ee83addf503c598ee3a80ddf3d",
+                "sha256:255a35bf29185f44b412e31a927d9dcedda7c2c380127ecc4fbf2f61b72fa978",
+                "sha256:2dbfbbded947a83a3dffc2bd1ec4750c17e40904692186e2c55a3ad314ca0222",
+                "sha256:2e92aa32300a0b5e4175caec7769f482b292769807024a86d674b3f19b8e3755",
+                "sha256:316c1b8723afa9870567cd6dff35d440b2afeda53aa13da6c5ab85f98ed6f5ca",
+                "sha256:333bfad77aa9cd11711febfb75eed0bb537a1d022e1c252714dad38993590240",
+                "sha256:39dafa2eaf577d1969f289dc9a44501859a1897eb45bd589e93ce843fc610800",
+                "sha256:3ce83f17f641a62a4dfb0ba1b8a3c1ced7c842f511b5450d90c030c7828e3693",
+                "sha256:46d5ec90276f71af3a29917b30f2aec2315a2759b5f8d45b3b63a07ca8a070a3",
+                "sha256:48d5bc80ab0af6b60c4163c5617f5cd23f2f880d7600940870ea5055816af024",
+                "sha256:4ba0def4abef058c0e5101e05e3d5266e6fffb9795bbf8be0fe912a7361a0209",
+                "sha256:5af390fa9faf56c93252dab09ea57cd020c9123aa921b63a0ed51832fdb492e7",
+                "sha256:5e574664f1468872cd40f74e4811e22b1aa4de9399d6bcfdf1ee6ea94c017fcf",
+                "sha256:625befa3bc9b40746a749115cc6a15bf20b9bd7597ca55d646205b479a2c99c7",
+                "sha256:6261bee7c5abadeac7497f8f1c43e521da78dd13b0a2439f526a7b0fc3788824",
+                "sha256:657ad80de8ec9ed656f28844efc801a0802961e8c6a85038d97ff6f555ef4919",
+                "sha256:6b89dc51206e4971c5568c797991eaaef5dc2a6118d67165858ad11752dba055",
+                "sha256:6e66780f14c2efaf989cd3ac613b03ee6a8e3a0ba7b96c0bb14adca71a427e55",
+                "sha256:6fb3f85870ae26896bb44e67db94045f2ebf00c5d41e6b66cdcbb5afd644fc18",
+                "sha256:701e08457183da70ed96b35a6b43e6ba1df0b47c837b063cde39a1fbe1aeda81",
+                "sha256:70761fd3c576b027eec882b43ee0a8e5b22ff9c20cdf4d0400e104bc29e53e34",
+                "sha256:73b400fdc22de84bae0dbf1a22613928a41612ec0a3d6ed47caf7ad4d3d0f2ff",
+                "sha256:7412a36798966624dc4c57d64aa43c2d1100b348abd98daaac8e99e57d87e1d7",
+                "sha256:78ecb8d42f50d393af912bfb1fb1dcc9aabe9967973efb49ee577e8f1cea494c",
+                "sha256:7c6a9948916a7bbcc6d3a9f6fb75db1acb5546078023bfb3db6efabcd5a67527",
+                "sha256:7c72d08acdf573455b2b9d2b75b8237654841d63a48bc2327dc102c6ee89b75a",
+                "sha256:7d98ce3c42921bb91566121b658e0d9d59a9082a9bd6f473190607ff25ab637f",
+                "sha256:845a8b83798b2fb11b09928413cb32692866bfbc28830a433d9fa4c8c3720dd0",
+                "sha256:94d38eba4d1b5eb3e6bfece0651b855a35c44f32fd91f512ab4ba41b8c0d3e66",
+                "sha256:9a13661681d17e43009bb3e85e837aa1ec5feeea1e3654682a01b8821940f8b3",
+                "sha256:a0e5dff6701fa615f165306e642709e1c1550d5b237c5a7a6ea299886828bd50",
+                "sha256:a2239556ff7241584ce57be1facf25081669bb457a9e5cbe68cce4aae6567aa1",
+                "sha256:a325600c83e61e3c9cebc0c2b1c8c4140fa887f789085075e8f44c8ff2547eb9",
+                "sha256:a3566acfbcde46911c52810374ecc0354fdb841284a3efef6ff7105bc007e9a8",
+                "sha256:a634a4730ce0b0934ed75e45beba730968e12b4dafbb22f69b3b2f616d9e644e",
+                "sha256:a6d055f01b83b1a4df8bb0c61983d3bdffa913764488910af3620e5c2450bf83",
+                "sha256:a752ecd1a26000a6d67be7c9a2e93801994a8b3f866ac95b672fbc00225ca91a",
+                "sha256:a9ba2a63777027b06b116e1ea8248e66fd1bedc2c644f93124b81a91ddbf6d88",
+                "sha256:aaa038eafb7186a4abbb311fcf20724be9363645882bbce540bef4797e812a7a",
+                "sha256:af586e85144023686fb0af09c8cdf672484ea182f352e7ceead3d832de381e1b",
+                "sha256:b0a0cf39f589e52d801fdef418305562bc030cdf8929217463c8433c65fd5c2f",
+                "sha256:b1c4874331ab960429caca81acb9d2932170d66d6d6f87e65dc4507a85aca152",
+                "sha256:b3b5b3cbc3fdf4fcfa292529df2a85b5d9c7053913a739d3069af1e12e12219f",
+                "sha256:b542d56ed1b8d5cf3bb36326f814bd2fbe8812dfd2582b80a15689ea433c0e35",
+                "sha256:b6ea08758b6673610b3c5bdf47189286cf9c58b1077558706a2f6f8744922527",
+                "sha256:b754240daafecd9d5fce426b0fbaaed03f4ebb130745c8a4ae9231fffb8d75e5",
+                "sha256:b772bab31cbd9cb911e41e1a611ebc9497f9a32a7348e2747c38210f75c00f41",
+                "sha256:b88d1742159bc93a078733f9789f563cef26f5e370eba810476a71aa98e5fbc2",
+                "sha256:b8bf42d3b32f586f4c9e37541769993783a534ad35531ce8a4379f6fa664fba9",
+                "sha256:bc9ac81e73573516070d24ce15da91281922811f385645df32bd3c8a45ab4684",
+                "sha256:c188db6cf9e14dbbb42f5254292be96f05374a35e7dfa087cc2140f0ff4f10f6",
+                "sha256:c55782a55f4a013a78ac5b6ee4b8731a192dea7ab09f1b6b3044c96d5128edd4",
+                "sha256:c5cab230e7cabdae9ff23c12271231283efefb944c1b79bed79a91beb65ba547",
+                "sha256:cbf8672edeb7b7128c4a939274801f0e32bbf5159987815e3d1eace625264a46",
+                "sha256:cc2894fe91f31a513860238ede69fe47fada21f9e7ddfe73f7f9fef93a971e41",
+                "sha256:cda9e628b1315beec8341e8c04aac9a0b910650b05e0751e42e399d5694aeacb",
+                "sha256:ceae3ab9e11a27aaab42878f1d203600dfd24f0e43678b47298219a0f10c0d30",
+                "sha256:ced944dcdd561476deef7cb7bfd4987c69fffbfeff6d02ca4d5d4fd592d559b7",
+                "sha256:d04ca462cb99077e6c059e97c072957caf2918e6e4191e3161c01c439e0193de",
+                "sha256:d1131562ddc2ea8a446f66c2648d7dabec2b3816fc818528eb978a75a6d23b2e",
+                "sha256:d1740776b70367277323fafb76bcf09753a5cc9824f5d705bac22a34ff3668ea",
+                "sha256:d6e11ffd43184d529d6752d6dcb62b994f903038a17ea2168ef1910c96324d26",
+                "sha256:d73e10772152605f6648ba4410318594f1043bbfe36d2fadee7c4b8912eff7c5",
+                "sha256:da8288bc4a7807c6715416deed1c57d94d5e03e93537889e002bf985be503f1a",
+                "sha256:db93608a246da44d728842b8fa9e45aa9782db76955f634a707739a8d53ff544",
+                "sha256:dcd3d0009fbb6e454d729f8b22d0063bd9171c31a55e0f0271119bd4f2700023",
+                "sha256:dd1f49f949a658c4e8f81ed73f9aad25fcc7d4f62f767f591e749e30038c4e1d",
+                "sha256:dd6ff2192f34bd622883c745a56f492b1c9ccd44e14953e8051c33024a2947d5",
+                "sha256:e018a4921657c2d3f89c720b7b90b9182e277178a04a7e9542cc79d7d787ca51",
+                "sha256:e2b7670c0c8c6b501464150dd49dd0d6be6cb7f049e064124911cec5514fa19e",
+                "sha256:e7a33322e08021c37e89cae8ff06327503e8a1719e97c69f32c31cbf6c30d72c",
+                "sha256:e8a82e35d52ad6f867e88096a1a2b9bdc7ec4d5e65c7b4976a248bf2d1a32a93",
+                "sha256:e9faf8d4712d5ea301d74abfcf6dafe4b7f4af7936e91f283b0ad7bf69ed3e3a",
+                "sha256:ec5ca7c0007ce268048bbe0ffc6846ed1616cf3d8628b136e81d5e64ff3f52a2",
+                "sha256:eee42a1cc06565f6b21caa1f504ec15e07de7ebfd520ab57f8cb3308bc118e22",
+                "sha256:f2acf9bbcd514e901f82c4ca6926bbd2ae61716728f110b4343eb0a69612d018",
+                "sha256:f55c1ddcc1f6050b07d468ce594f55dbf6107b459e16f735d26818d7be1e9538",
+                "sha256:f6977a520bd96e097c8a37a8cbb9faa1ea99d21bf84190195056e25f688af73d",
+                "sha256:f94c7d22fb36b184734dded7345a04ec5f95130421c775b8b0c65044ef073f34",
+                "sha256:fa8957e9a1b202cb45e6b839c241cd986c897be1e722b81d2f32e9c6aeee80b0",
+                "sha256:fd3854148005c808c485c754a184c71116372263709958b42aefbef2e5dd373a",
+                "sha256:fe5872ce6f9627deac8314bdffd3862624227c3de4c17ef0cc78bbf0402999eb",
+                "sha256:ffbae429ba9e42d0582d3ac63fdb410338892468a2107d8ff68228ec9a39a0ed"
             ],
             "index": "pypi",
-            "version": "==4.1.1"
+            "version": "==3.12"
         },
         "pyparsing": {
             "hashes": [
@@ -1915,7 +1922,7 @@
                 "sha256:f87cc2863ef33c709e237d4b5f4502a62a00fab450c9e020892e8e2ede5847f5",
                 "sha256:fd8da6d0124efa2f67d86fa70c851022f87c98e205f0594e1fae044e7119a5a6"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.18.1"
         },
         "pytest": {
@@ -1923,7 +1930,7 @@
                 "sha256:841132caef6b1ad17a9afde46dc4f6cfa59a05f9555aae5151f73bdf2820ca63",
                 "sha256:92f723789a8fdd7180b6b06483874feca4c48a5c76968e03bb3e7f806a1869ea"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==7.1.1"
         },
         "pytest-factoryboy": {
@@ -2177,7 +2184,7 @@
                 "sha256:ff3fa8ea0e09e38677762afc6e14cad77b5e125b0ea70c9bba1992f02c93b028",
                 "sha256:ff746a69ff2ef25f62b36338c615dd15954ddc3ab8e73530237dd73235e76d62"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==1.0.2"
         },
         "scipy": {
@@ -2212,7 +2219,7 @@
                 "sha256:f7eaea089345a35130bc9a39b89ec1ff69c208efa97b3f8b25ea5d4c41d88094",
                 "sha256:f99d206db1f1ae735a8192ab93bd6028f3a42f6fa08467d37a14eb96c9dd34a3"
             ],
-            "markers": "python_version < '3.11' and python_full_version >= '3.7.0'",
+            "markers": "python_version < '3.11' and python_version >= '3.7'",
             "version": "==1.7.3"
         },
         "send2trash": {
@@ -2238,7 +2245,7 @@
                 "sha256:26ead7d1f93efc0f8c804d9fafafbe4a44b179580a7105754b245155f9af05a8",
                 "sha256:47c7b0c0f8fc10eec4cf1e71c6fdadf8decaa74ffa087e68cd1c20db7ad6a592"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==62.1.0"
         },
         "six": {
@@ -2307,7 +2314,7 @@
                 "sha256:874d4ea3183536c1782d13c7c91342ef0cf4e5ee1d53633029cbc972c8760bd8",
                 "sha256:94d1cfab63525993f7d5c9b469a50a18d0cdf39435b59785715539dd41e36c0d"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.13.3"
         },
         "testpath": {
@@ -2331,7 +2338,7 @@
                 "sha256:939de3e7a6161af0c887ef91b7d41a53e7c5a1ca976325f429cb46ea9bc30ecc",
                 "sha256:de526c12914f0c550d15924c62d72abc48d6fe7364aa87328337a31007fe8a4f"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==2.0.1"
         },
         "torch": {
@@ -2444,7 +2451,7 @@
                 "sha256:059f456c5a7c1c82b98c2e8c799f39c9b8128f6d0d46941ee118daace9eb70c7",
                 "sha256:2d313cc50a42cd6c277e7d7dc8d4d7fedd06a2c215f78766ae7b1a66277e0033"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==5.1.1"
         },
         "typing-extensions": {
@@ -2452,7 +2459,7 @@
                 "sha256:6657594ee297170d19f67d55c05852a874e7eb634f4f753dbd667855e07c1708",
                 "sha256:f1c24655a0da0d1b67f07e17a5e6b2a105894e6824b92096378bb3668ef02376"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==4.2.0"
         },
         "urllib3": {
@@ -2460,7 +2467,7 @@
                 "sha256:44ece4d53fb1706f667c9bd1c648f5469a2ec925fcf3a776667042d645472c14",
                 "sha256:aabaf16477806a5e1dd19aa41f8c2b7950dd3c746362d7e3223dbe6de6ac448e"
             ],
-            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' and python_version < '4.0'",
+            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' and python_full_version < '4.0.0'",
             "version": "==1.26.9"
         },
         "vine": {
@@ -2490,7 +2497,7 @@
                 "sha256:50b21db0058f7a953d67cc0445be4b948d7fc196ecbeb8083d68d94628e4abf6",
                 "sha256:722b171be00f2b90e1d4fb2f2b53146a536ca38db1da8ff49c972a4e1365d0ef"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==1.3.2"
         },
         "werkzeug": {
@@ -2587,7 +2594,7 @@
                 "sha256:6b351bbb12dd58af57ffef05bc78425d08d1914e0fd68ee14143b7ade023c5bc",
                 "sha256:837f2f0e0ca79481b92884962b914eba4e72b7a2daaf1f939c890ed0124b834b"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==3.0.1"
         },
         "yarl": {
@@ -2673,7 +2680,7 @@
                 "sha256:56bf8aadb83c24db6c4b577e13de374ccfb67da2078beba1d037c17980bf43ad",
                 "sha256:c4f6e5bbf48e74f7a38e7cc5b0480ff42b0ae5178957d564d18932525d5cf099"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==3.8.0"
         }
     },
@@ -2693,14 +2700,6 @@
             ],
             "version": "==1.4.4"
         },
-        "appnope": {
-            "hashes": [
-                "sha256:02bd91c4de869fbb1e1c50aafc4098827a7a54ab2f39d9dcba6c9547ed920e24",
-                "sha256:265a455292d0bd8a72453494fa24df5a11eb18373a60c7c0430889f22548605e"
-            ],
-            "markers": "platform_system == 'Darwin'",
-            "version": "==0.1.3"
-        },
         "argon2-cffi": {
             "hashes": [
                 "sha256:8c976986f2c5c0e5000919e6de187906cfd81fb1c72bf9d88c01177e77da7f80",
@@ -2787,7 +2786,7 @@
                 "sha256:08a1fe86d253b5c88c92cc3d810fd8048a16d15762e1e5b74d502256e5926aa1",
                 "sha256:c6d6cc054bdc9c83b48b8083e236e5f00f238428666d2ce2e083eaa5fd568565"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==5.0.0"
         },
         "certifi": {
@@ -2897,7 +2896,7 @@
                 "sha256:e3513399177dd37af4c1332df52da5da1d0c387e5927dc4c0709e26ee7302e8f",
                 "sha256:eb1946efac0c0c3d411cea0b5ac772fbde744109fd9520fb0c5a51979faf05ad"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==1.6.0"
         },
         "decorator": {
@@ -2975,7 +2974,7 @@
                 "sha256:9cd540a9352e432c7246a48fe4e8712b10acb1df2ad1f30e8c070b82ae1fed85",
                 "sha256:f8314284bfffbdcfa0ff3d7992b023d4c628ced6feb957351d4c48d059f56bc0"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==3.6.0"
         },
         "graphviz": {
@@ -3038,7 +3037,7 @@
                 "sha256:0e28273e290858393e86e152b104e5506a79c13d25b951ac6eca220051b4be60",
                 "sha256:2b0987af43c0d4b62cecb13c592755f599f96f29aafe36c01731aaa96df30d39"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==6.13.0"
         },
         "ipytest": {
@@ -3054,7 +3053,7 @@
                 "sha256:468abefc45c15419e3c8e8c0a6a5c115b2127bafa34d7c641b1d443658793909",
                 "sha256:86df2cf291c6c70b5be6a7b608650420e89180c8ec74f376a34e2dc15c3400e7"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==7.32.0"
         },
         "ipython-genutils": {
@@ -3107,7 +3106,7 @@
                 "sha256:636694eb41b3535ed608fe04129f26542b59ed99808b4f688aa32dcf55317a83",
                 "sha256:77281a1f71684953ee8b3d488371b162419767973789272434bbc3f29d9c8823"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==4.4.0"
         },
         "jupyter": {
@@ -3124,7 +3123,7 @@
                 "sha256:44045448eadc12493d819d965eb1dc9d10d1927698adbb9b14eb9a3a4a45ba53",
                 "sha256:8fdbad344a8baa6a413d86d25bbf87ce21cb2b4aa5a8e0413863b9754eb8eb8a"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==7.2.2"
         },
         "jupyter-console": {
@@ -3140,7 +3139,7 @@
                 "sha256:a6de44b16b7b31d7271130c71a6792c4040f077011961138afed5e5e73181aec",
                 "sha256:e7f5212177af7ab34179690140f188aa9bf3d322d8155ed972cbded19f55b6f3"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==4.10.0"
         },
         "jupyter-server": {
@@ -3148,7 +3147,7 @@
                 "sha256:72dd1ff5373d2def94e80632ba4397e504cc9200c5b5f44b5b0af2e062a73353",
                 "sha256:c756f87ad64b84e2aa522ef482445e1a93f7fe4a5fc78358f4636e53c9a0463a"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==1.16.0"
         },
         "jupyterlab": {
@@ -3164,7 +3163,7 @@
                 "sha256:2405800db07c9f770863bcf8049a529c3dd4d3e28536638bd7c1c01d2748309f",
                 "sha256:7405d7fde60819d905a9fa8ce89e4cd830e318cdad22a0030f7a901da705585d"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.2.2"
         },
         "jupyterlab-server": {
@@ -3344,7 +3343,7 @@
                 "sha256:36dbaa88ffaf5dc05d149deb97504b86ba648f4a80a60b8a58ac94acab2daeb5",
                 "sha256:89184baa2d66b8ac3c8d3df57cbcf16f34148954d410a2fb3e897d7c18f2479d"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.3.7"
         },
         "nbclient": {
@@ -3352,7 +3351,7 @@
                 "sha256:40c52c9b5e3c31faecaee69f202b3f53e38d7c1c563de0fadde9d7eda0fdafe8",
                 "sha256:47ac905af59379913c1f8f541098d2550153cf8dc58553cbe18c702b181518b0"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.5.13"
         },
         "nbconvert": {
@@ -3360,7 +3359,7 @@
                 "sha256:21163a8e2073c07109ca8f398836e45efdba2aacea68d6f75a8a545fef070d4e",
                 "sha256:e01d219f55cc79f9701c834d605e8aa3acf35725345d3942e3983937f368ce14"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==6.4.5"
         },
         "nbformat": {
@@ -3368,7 +3367,7 @@
                 "sha256:38856d97de49e8292e2d5d8f595e9d26f02abfd87e075d450af4511870b40538",
                 "sha256:fcc5ab8cb74e20b19570b5be809e2dba9b82836fd2761a89066ad43394ba29f5"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==5.3.0"
         },
         "nest-asyncio": {
@@ -3391,7 +3390,7 @@
                 "sha256:709b1856a564fe53054796c80e17a67262071c86bfbdfa6b96aaa346113c555a",
                 "sha256:b4a6baf2eba21ce67a0ca11a793d1781b06b8078f34d06c710742e55f3eee505"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==6.4.11"
         },
         "notebook-shim": {
@@ -3399,7 +3398,7 @@
                 "sha256:02432d55a01139ac16e2100888aa2b56c614720cec73a27e71f40a5387e45324",
                 "sha256:7897e47a36d92248925a2143e3596f19c60597708f7bef50d81fcd31d7263e85"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.1.0"
         },
         "packaging": {
@@ -3597,7 +3596,7 @@
                 "sha256:f87cc2863ef33c709e237d4b5f4502a62a00fab450c9e020892e8e2ede5847f5",
                 "sha256:fd8da6d0124efa2f67d86fa70c851022f87c98e205f0594e1fae044e7119a5a6"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.18.1"
         },
         "pytest": {
@@ -3605,7 +3604,7 @@
                 "sha256:841132caef6b1ad17a9afde46dc4f6cfa59a05f9555aae5151f73bdf2820ca63",
                 "sha256:92f723789a8fdd7180b6b06483874feca4c48a5c76968e03bb3e7f806a1869ea"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==7.1.1"
         },
         "pytest-factoryboy": {
@@ -3861,7 +3860,7 @@
                 "sha256:26ead7d1f93efc0f8c804d9fafafbe4a44b179580a7105754b245155f9af05a8",
                 "sha256:47c7b0c0f8fc10eec4cf1e71c6fdadf8decaa74ffa087e68cd1c20db7ad6a592"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==62.1.0"
         },
         "six": {
@@ -3893,7 +3892,7 @@
                 "sha256:874d4ea3183536c1782d13c7c91342ef0cf4e5ee1d53633029cbc972c8760bd8",
                 "sha256:94d1cfab63525993f7d5c9b469a50a18d0cdf39435b59785715539dd41e36c0d"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==0.13.3"
         },
         "tinycss2": {
@@ -3917,7 +3916,7 @@
                 "sha256:939de3e7a6161af0c887ef91b7d41a53e7c5a1ca976325f429cb46ea9bc30ecc",
                 "sha256:de526c12914f0c550d15924c62d72abc48d6fe7364aa87328337a31007fe8a4f"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==2.0.1"
         },
         "tornado": {
@@ -3972,7 +3971,7 @@
                 "sha256:059f456c5a7c1c82b98c2e8c799f39c9b8128f6d0d46941ee118daace9eb70c7",
                 "sha256:2d313cc50a42cd6c277e7d7dc8d4d7fedd06a2c215f78766ae7b1a66277e0033"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==5.1.1"
         },
         "typed-ast": {
@@ -4010,7 +4009,7 @@
                 "sha256:6657594ee297170d19f67d55c05852a874e7eb634f4f753dbd667855e07c1708",
                 "sha256:f1c24655a0da0d1b67f07e17a5e6b2a105894e6824b92096378bb3668ef02376"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==4.2.0"
         },
         "urllib3": {
@@ -4018,7 +4017,7 @@
                 "sha256:44ece4d53fb1706f667c9bd1c648f5469a2ec925fcf3a776667042d645472c14",
                 "sha256:aabaf16477806a5e1dd19aa41f8c2b7950dd3c746362d7e3223dbe6de6ac448e"
             ],
-            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' and python_version < '4.0'",
+            "markers": "python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' and python_full_version < '4.0.0'",
             "version": "==1.26.9"
         },
         "virtualenv": {
@@ -4048,7 +4047,7 @@
                 "sha256:50b21db0058f7a953d67cc0445be4b948d7fc196ecbeb8083d68d94628e4abf6",
                 "sha256:722b171be00f2b90e1d4fb2f2b53146a536ca38db1da8ff49c972a4e1365d0ef"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==1.3.2"
         },
         "widgetsnbextension": {
@@ -4133,7 +4132,7 @@
                 "sha256:56bf8aadb83c24db6c4b577e13de374ccfb67da2078beba1d037c17980bf43ad",
                 "sha256:c4f6e5bbf48e74f7a38e7cc5b0480ff42b0ae5178957d564d18932525d5cf099"
             ],
-            "markers": "python_full_version >= '3.7.0'",
+            "markers": "python_version >= '3.7'",
             "version": "==3.8.0"
         }
     }
diff --git a/recommender/api/schemas/recommendation.py b/recommender/api/schemas/recommendation.py
index 66568ea7..d94c8515 100644
--- a/recommender/api/schemas/recommendation.py
+++ b/recommender/api/schemas/recommendation.py
@@ -92,6 +92,15 @@
             description="Version of the recommendation engine",
             example="RL",
         ),
+        "elastic_services": fields.List(
+            fields.Integer(
+                title="Service ID", description="The unique identifier of the service"
+            ),
+            required=True,
+            title="ElasticSearch services list",
+            description="List of services IDs from ElasticSearch",
+            example=[1, 2, 3, 4],
+        ),
         "search_data": fields.Nested(search_data, required=True),
     },
 )
diff --git a/recommender/engines/base/base_inference_component.py b/recommender/engines/base/base_inference_component.py
index 64499bb6..016e755e 100644
--- a/recommender/engines/base/base_inference_component.py
+++ b/recommender/engines/base/base_inference_component.py
@@ -5,7 +5,7 @@
 
 from abc import ABC, abstractmethod
 import random
-from typing import Dict, Any, List
+from typing import Dict, Any, List, Tuple
 
 from recommender.engines.panel_id_to_services_number_mapping import PANEL_ID_TO_K
 from recommender.errors import (
@@ -46,11 +46,13 @@ def __call__(self, context: Dict[str, Any]) -> List[int]:
         """
 
         user = self._get_user(context)
-        search_data = self._get_search_data(context)
+        elastic_services, search_data = self._get_elastic_services_and_search_data(
+            context
+        )
 
         if user is not None:
-            return self._for_logged_user(user, search_data)
-        return self._for_anonymous_user(search_data)
+            return self._for_logged_user(user, elastic_services, search_data)
+        return self._for_anonymous_user(elastic_services)
 
     @abstractmethod
     def _load_models(self) -> None:
@@ -76,26 +78,22 @@ def _get_user(self, context: Dict[str, Any]) -> User:
 
         return user
 
-    def _get_search_data(self, context: Dict[str, Any]) -> SearchData:
-        search_data = context.get("search_data")
-        search_data.pop(
-            "rating", None
-        )  # We don't and we shouldn't take rating into consideration
-
-        # To prevent q being None (for SearchPhraseEncoder it must be a string)
-        search_data["q"] = search_data.get("q", "")
-
-        search_data = SearchData(**search_data)
-
-        return search_data
+    @staticmethod
+    def _get_elastic_services_and_search_data(
+        context: Dict[str, Any]
+    ) -> [Tuple[int], SearchData]:
+        return tuple(context.get("elastic_services")), context.get("search_data")
 
     @abstractmethod
-    def _for_logged_user(self, user: User, search_data: SearchData) -> List[int]:
+    def _for_logged_user(
+        self, user: User, elastic_services: Tuple[int], search_data: SearchData
+    ) -> List[int]:
         """
         Generate recommendation for logged user
 
         Args:
             user: user for whom recommendation will be generated.
+            elastic_services: item space from the Marketplace.
             search_data: search phrase and filters information for narrowing
              down an item space.
 
@@ -103,19 +101,20 @@ def _for_logged_user(self, user: User, search_data: SearchData) -> List[int]:
             recommended_services_ids: List of recommended services ids.
         """
 
-    def _for_anonymous_user(self, search_data: SearchData) -> List[int]:
+    def _for_anonymous_user(self, elastic_services: Tuple[int]) -> List[int]:
         """
         Generate recommendation for anonymous user
 
         Args:
-            search_data: search phrase and filters information for narrowing
-             down an item space.
+            elastic_services: item space from the Marketplace.
 
         Returns:
             recommended_services_ids: List of recommended services ids.
         """
 
-        candidate_services = list(retrieve_services_for_recommendation(search_data))
+        candidate_services = list(
+            retrieve_services_for_recommendation(elastic_services)
+        )
         if len(candidate_services) < self.K:
             raise InsufficientRecommendationSpaceError()
         recommended_services = random.sample(list(candidate_services), self.K)
diff --git a/recommender/engines/ncf/inference/ncf_inference_component.py b/recommender/engines/ncf/inference/ncf_inference_component.py
index 29731ffc..4ae8116f 100644
--- a/recommender/engines/ncf/inference/ncf_inference_component.py
+++ b/recommender/engines/ncf/inference/ncf_inference_component.py
@@ -3,7 +3,7 @@
 
 """Neural Collaborative Filtering Inference Component"""
 
-from typing import List
+from typing import List, Tuple
 
 import torch
 
@@ -73,11 +73,14 @@ def user_and_services_to_tensors(user, services):
 
         return users_ids, users_tensor, services_ids, services_tensor
 
-    def _for_logged_user(self, user: User, search_data: SearchData) -> List[int]:
+    def _for_logged_user(
+        self, user: User, elastic_services: Tuple[int], search_data: SearchData
+    ) -> List[int]:
         """Generate recommendation for logged user.
 
         Args:
             user: user for whom recommendation will be generated.
+            elastic_services: item space from the Marketplace.
             search_data: search phrase and filters information for narrowing
              down an item space.
 
@@ -86,7 +89,9 @@ def _for_logged_user(self, user: User, search_data: SearchData) -> List[int]:
         """
 
         candidate_services = list(
-            retrieve_services_for_recommendation(search_data, user.accessed_services)
+            retrieve_services_for_recommendation(
+                elastic_services, user.accessed_services
+            )
         )
         if len(candidate_services) < self.K:
             raise InsufficientRecommendationSpaceError()
diff --git a/recommender/engines/rl/inference/rl_inference_component.py b/recommender/engines/rl/inference/rl_inference_component.py
index a9ed163d..919bd36a 100644
--- a/recommender/engines/rl/inference/rl_inference_component.py
+++ b/recommender/engines/rl/inference/rl_inference_component.py
@@ -37,15 +37,18 @@ def _load_models(self) -> None:
         )
         self.service_selector = ServiceSelector(self.service_embedder)
 
-    def _for_logged_user(self, user: User, search_data: SearchData) -> Tuple[int]:
-        state = create_state(user, search_data)
+    def _for_logged_user(
+        self, user: User, elastic_services: Tuple[int], search_data: SearchData
+    ) -> Tuple[int]:
+        state = create_state(user, elastic_services, search_data)
         state_tensors = self.state_encoder([state])
         weights_tensor = self._get_weights(state_tensors)
-        search_data_mask = self._get_search_data_mask(state_tensors)
-        service_ids = self.service_selector(weights_tensor, search_data_mask)
+        services_mask = self._get_service_mask(state_tensors)
+        service_ids = self.service_selector(weights_tensor, services_mask)
         return service_ids
 
-    def _get_search_data_mask(self, state_tensors):
+    @staticmethod
+    def _get_service_mask(state_tensors):
         return state_tensors[2][0]
 
     def _get_weights(self, state_tensors):
diff --git a/recommender/engines/rl/ml_components/sars_encoder.py b/recommender/engines/rl/ml_components/sars_encoder.py
index fc770709..8542f624 100644
--- a/recommender/engines/rl/ml_components/sars_encoder.py
+++ b/recommender/engines/rl/ml_components/sars_encoder.py
@@ -62,8 +62,8 @@ def __call__(
         states = [SARS.state for SARS in SARSes]
         next_states = [SARS.next_state for SARS in SARSes]
         all_states = states + next_states
-        all_states_tensors = self.state_encoder(all_states)
 
+        all_states_tensors = self.state_encoder(all_states, verbose=True)
         states_tensors = tuple(t[: len(states)] for t in all_states_tensors)
         next_states_tensors = tuple(t[len(states) :] for t in all_states_tensors)
 
diff --git a/recommender/engines/rl/ml_components/sarses_generator.py b/recommender/engines/rl/ml_components/sarses_generator.py
index 295a7fec..d06a4474 100644
--- a/recommender/engines/rl/ml_components/sarses_generator.py
+++ b/recommender/engines/rl/ml_components/sarses_generator.py
@@ -17,6 +17,7 @@
 import itertools
 import time
 from typing import List, Union
+from copy import deepcopy
 import billiard
 import torch
 from mongoengine import DoesNotExist
@@ -252,6 +253,7 @@ def skipable_func(*args, **kwargs):
 def generate_sars(recommendation, root_uas):
     """Generate sars for given recommendation and root user actions"""
 
+    rec = deepcopy(recommendation)  # elastic_services memory leak
     user = recommendation.user or _get_empty_user()
 
     # Create reward
@@ -266,6 +268,7 @@ def generate_sars(recommendation, root_uas):
     state = State(
         user=user,
         services_history=services_history_before,
+        elastic_services=rec.elastic_services,
         search_data=recommendation.search_data,
     ).save()
 
@@ -283,6 +286,7 @@ def generate_sars(recommendation, root_uas):
     next_state = State(
         user=user,
         services_history=services_history_after,
+        elastic_services=rec.elastic_services,
         search_data=next_recommendation.search_data,
     ).save()
 
diff --git a/recommender/engines/rl/ml_components/search_data_encoder.py b/recommender/engines/rl/ml_components/search_data_encoder.py
deleted file mode 100644
index da935631..00000000
--- a/recommender/engines/rl/ml_components/search_data_encoder.py
+++ /dev/null
@@ -1,57 +0,0 @@
-# pylint: disable=missing-module-docstring, no-member, too-few-public-methods, invalid-name
-from typing import List
-
-import torch
-
-from recommender.engines.rl.utils import (
-    create_index_id_map,
-    get_service_indices,
-)
-from recommender.models import SearchData, User, Service
-from recommender.services.fts import retrieve_forbidden_services, filter_services
-from recommender.engines.rl.ml_components.services_history_generator import (
-    get_ordered_services,
-)
-from recommender.errors import SizeOfUsersAndSearchDataError
-
-
-class SearchDataEncoder:
-    """Encodes search data and user's ordered services into a binary mask"""
-
-    def __init__(self):
-        all_services = list(Service.objects.order_by("id"))
-        self.I = len(all_services)
-        self.index_id_map = create_index_id_map(all_services)
-
-        forbidden_services = retrieve_forbidden_services()
-        self.forbidden_service_indices = get_service_indices(
-            self.index_id_map, forbidden_services.distinct("id")
-        )
-
-    def __call__(
-        self, users: List[User], search_data: List[SearchData]
-    ) -> torch.Tensor:
-        if not len(users) == len(search_data):
-            raise SizeOfUsersAndSearchDataError()
-
-        batch_size = len(users)
-
-        mask = torch.zeros(batch_size, self.I)
-
-        for i in range(batch_size):
-            filtered_services = filter_services(search_data[i])
-            ordered_services = get_ordered_services(users[i])
-
-            filtered_service_indices = get_service_indices(
-                self.index_id_map, filtered_services.distinct("id")
-            )
-
-            ordered_service_indices = get_service_indices(
-                self.index_id_map, [s.id for s in ordered_services]
-            )
-
-            mask[i, filtered_service_indices] = 1
-            mask[i, ordered_service_indices] = 0
-            mask[i, self.forbidden_service_indices] = 0
-
-        return mask
diff --git a/recommender/engines/rl/ml_components/service_encoder.py b/recommender/engines/rl/ml_components/service_encoder.py
new file mode 100644
index 00000000..a51f6d1b
--- /dev/null
+++ b/recommender/engines/rl/ml_components/service_encoder.py
@@ -0,0 +1,84 @@
+# pylint: disable=missing-module-docstring, no-member, too-few-public-methods, invalid-name, line-too-long
+from typing import List
+from copy import deepcopy
+import torch
+from tqdm import tqdm
+from recommender.engines.rl.utils import (
+    create_index_id_map,
+    get_service_indices,
+)
+from recommender.models import User, Service, State
+from recommender.services.fts import retrieve_forbidden_services
+from recommender.engines.rl.ml_components.services_history_generator import (
+    get_ordered_services,
+)
+from recommender.errors import SizeOfUsersAndElasticServicesError
+from logger_config import get_logger
+
+logger = get_logger(__name__)
+
+
+class ServiceEncoder:
+    """Encodes search data and user's ordered services into a binary mask"""
+
+    def __init__(self):
+        all_services = list(Service.objects.order_by("id"))
+        self.I = len(all_services)
+        self.index_id_map = create_index_id_map(all_services)
+
+        forbidden_services = retrieve_forbidden_services()
+        self.forbidden_service_indices = get_service_indices(
+            self.index_id_map, forbidden_services.distinct("id")
+        )
+
+    def __call__(
+        self, users: List[User], states: List[State], verbose=False
+    ) -> torch.Tensor:
+        batch_size = len(users)
+        states_len = len(states)
+
+        if batch_size != states_len:
+            # Each state has elastic_services
+            raise SizeOfUsersAndElasticServicesError
+
+        mask = torch.zeros(batch_size, self.I)
+
+        if verbose:
+            logger.info("Creating mask...")
+
+        for i in tqdm(range(batch_size), desc="States", disable=not verbose):
+            state = states.pop(0)
+
+            if states_len == 1:
+                # Model evaluation step - searchdata there is not saved, so it cannot be accessed
+                # This avoids mongodb ValidationError
+                state.search_data = None
+
+            # Just reading the tuple of references (which is the elastic_services)
+            # causes the IDs to be replaced with entire service objects in all referenced tuple.
+            # To avoid memory leak:
+            # - state has 3 references - direct manipulation on state object will result in memory leak
+            #   to avoid that, there is deepcopy made
+            state = deepcopy(state)
+            elastic_services = state.elastic_services
+            services_context = self.get_id_from_services(elastic_services)
+
+            ordered_services = get_ordered_services(users[i])
+
+            services_context_indices = get_service_indices(
+                self.index_id_map, services_context
+            )
+            ordered_service_indices = get_service_indices(
+                self.index_id_map, [s.id for s in ordered_services]
+            )
+
+            mask[i, services_context_indices] = 1
+            mask[i, ordered_service_indices] = 0
+            mask[i, self.forbidden_service_indices] = 0
+
+        return mask
+
+    @staticmethod
+    def get_id_from_services(services: List[Service]) -> List[int]:
+        """Get services IDs"""
+        return [service.id for service in services]
diff --git a/recommender/engines/rl/ml_components/state_encoder.py b/recommender/engines/rl/ml_components/state_encoder.py
index 47038134..dbbfd45d 100644
--- a/recommender/engines/rl/ml_components/state_encoder.py
+++ b/recommender/engines/rl/ml_components/state_encoder.py
@@ -4,16 +4,16 @@
 
 """Implementation of the State Encoder"""
 from typing import Tuple, List
-
 import torch
 from torch.nn.utils.rnn import pad_sequence
-
-from recommender.engines.rl.ml_components.search_data_encoder import (
-    SearchDataEncoder,
+from recommender.engines.rl.ml_components.service_encoder import (
+    ServiceEncoder,
 )
 from recommender.engines.autoencoders.ml_components.embedder import Embedder
-from recommender.models import Service
-from recommender.models import State
+from recommender.models import Service, State
+from logger_config import get_logger
+
+logger = get_logger(__name__)
 
 
 class StateEncoder:
@@ -30,9 +30,11 @@ def __init__(
         self.service_embedder = service_embedder
         self.use_cached_embeddings = use_cached_embeddings
         self.save_cached_embeddings = save_cached_embeddings
-        self.search_data_encoder = SearchDataEncoder()
+        self.service_encoder = ServiceEncoder()
 
-    def __call__(self, states: List[State]) -> Tuple[(torch.Tensor,) * 3]:
+    def __call__(
+        self, states: List[State], verbose: bool = False
+    ) -> Tuple[(torch.Tensor,) * 3]:
         """
         Encode given states to the tuple of tensors using appropriate encoders
          and embedders as follows:
@@ -43,8 +45,8 @@ def __call__(self, states: List[State]) -> Tuple[(torch.Tensor,) * 3]:
                 -> services_one_hot_tensors
                 --(service_embedder)--> services_dense_tensors
                 --(concat)--> services_histories_tensor
-            - state.search_data
-                --(search_data_encoder)--> search_data_tensor (mask)
+            - state.elastic_services
+                --(service encoder)--> service_tensor (mask)
 
 
         It makes batches from parts of states whenever it is possible to
@@ -52,13 +54,14 @@ def __call__(self, states: List[State]) -> Tuple[(torch.Tensor,) * 3]:
 
         Args:
             states: List of state objects. Each of them contains information
-             about user, user's services history and search data.
+             about user, user's services history and elastic services.
+            verbose: be verbose?
 
         Returns:
             Tuple of following tensors (in this order):
                 - user of shape [B, UE]
                 - service_histories_batch of shape [B, N, SE]
-                - search_data_mask of shape [B, I]
+                - service_mask of shape [B, I]
 
                 where:
                     - B is the batch_size and it is equal to len(states)
@@ -69,10 +72,14 @@ def __call__(self, states: List[State]) -> Tuple[(torch.Tensor,) * 3]:
                     - K is the first mask dim
                     - I is equal to itemspace size
         """
-
+        if verbose:
+            logger.info("Getting all users from states")
         users = [state.user for state in states]
+
+        if verbose:
+            logger.info("Getting all services_histories from states")
         services_histories = [state.services_history for state in states]
-        search_data_list = [state.search_data for state in states]
+
         with torch.no_grad():
             users_batch, _ = self.user_embedder(
                 users,
@@ -82,9 +89,10 @@ def __call__(self, states: List[State]) -> Tuple[(torch.Tensor,) * 3]:
             service_histories_batch = self._create_service_histories_batch(
                 services_histories
             )
-            search_data_masks_batch = self.search_data_encoder(users, search_data_list)
 
-        encoded_states = users_batch, service_histories_batch, search_data_masks_batch
+            service_masks_batch = self.service_encoder(users, states, verbose=verbose)
+
+        encoded_states = users_batch, service_histories_batch, service_masks_batch
 
         return encoded_states
 
diff --git a/recommender/engines/rl/ml_components/synthetic_dataset/dataset.py b/recommender/engines/rl/ml_components/synthetic_dataset/dataset.py
index 2c43f863..b383cb03 100644
--- a/recommender/engines/rl/ml_components/synthetic_dataset/dataset.py
+++ b/recommender/engines/rl/ml_components/synthetic_dataset/dataset.py
@@ -15,7 +15,7 @@
     NormalizationMode,
 )
 from recommender.models import Service, State, SearchData, Sars, User
-from recommender.services.fts import retrieve_services_for_recommendation
+from recommender.services.fts import retrieve_services_for_synthetic_sarses
 from recommender.engines.rl.ml_components.synthetic_dataset.rewards import (
     synthesize_reward,
     RewardGeneration,
@@ -23,9 +23,6 @@
 from recommender.engines.rl.ml_components.synthetic_dataset.service_engagement import (
     approx_service_engagement,
 )
-from recommender.engines.rl.ml_components.synthetic_dataset.users import (
-    synthesize_users,
-)
 
 
 def _normalize_embedded_services(embedded_services: torch.Tensor) -> torch.Tensor:
@@ -56,137 +53,23 @@ def _get_ordered_services(rewards_dict: Dict[Service, List[str]]) -> List[Servic
 def _get_relevant_search_data(user, ordered_services, k):
     for c in user.categories:
         search_data = SearchData(categories=[c])
-        if len(retrieve_services_for_recommendation(search_data, ordered_services)) > k:
+        if (
+            len(retrieve_services_for_synthetic_sarses(search_data, ordered_services))
+            > k
+        ):
             return search_data.save()
 
     for sd in user.scientific_domains:
         search_data = SearchData(scientific_domains=[sd])
-        if len(retrieve_services_for_recommendation(search_data, ordered_services)) > k:
+        if (
+            len(retrieve_services_for_synthetic_sarses(search_data, ordered_services))
+            > k
+        ):
             return search_data.save()
 
     return SearchData().save()
 
 
-def generate_dataset(
-    user_samples: int,
-    service_embedder: Embedder,
-    K: int = 3,
-    interactions_range: Tuple[int, int] = (3, 10),
-    reward_generation_mode: RewardGeneration = RewardGeneration.COMPLEX,
-    simple_reward_threshold: int = 0.5,
-    cluster_distributions: Tuple[(float,) * 7] = (
-        0.12,
-        0.2,
-        0.2,
-        0.12,
-        0.12,
-        0.12,
-        0.12,
-    ),  # experimentally chosen
-) -> Tuple[List[Sars], List[User]]:
-    """
-    Generates artificial SARS dataset that will be used for training
-    and benchmarking the RL agent.
-    It generates users belonging to a predefined category and scientific domain clusters
-    and for each one generates specified number of SARSes by:
-        - choosing an action randomly (LIRD is a off-policy RL algorithm)
-        - approximating heuristically the engagement of a user in a given service
-        - generates rewards based on this approximation for each service
-            (simulates the page transitions in the Marketplace portal)
-
-    Args:
-        user_samples: number of users to generate
-        K: number of recommended services in a single recommendation
-        interactions_range: range of recommendations
-            (as well as SARSes generated for each user)
-        reward_generation_mode: either "complex" or "simple". Specifies if
-            the simulated reward should be "complex" (drawn from the binomial
-            distribution + simulating user MP transitions) or "simple"
-            (high reward above a threshold - see "simple_reward_threshold" parameter)
-        simple_reward_threshold: defines the threshold above which
-            the simulated reward is high
-        cluster_distributions: defines the distribution of user clusters
-
-    Returns:
-        Generated SARSes and synthetic users
-    """
-
-    synthesized_users = synthesize_users(
-        user_samples, cluster_distributions=cluster_distributions
-    )
-
-    all_services = Service.objects.order_by("id")
-    service_embedded_tensors, index_id_map = service_embedder(all_services)
-
-    normalized_services = _normalize_embedded_services(service_embedded_tensors)
-    transition_rewards_df = pd.read_csv(TRANSITION_REWARDS_CSV_PATH, index_col="source")
-    sarses = []
-
-    for user in synthesized_users:
-        engaged_services = []
-        ordered_services = []
-
-        state = State(
-            user=user,
-            services_history=[],
-            search_data=_get_relevant_search_data(user, ordered_services, K),
-            synthetic=True,
-        ).save()
-
-        for _ in range(
-            np.random.randint(interactions_range[0], interactions_range[1] + 1)
-        ):
-            # Here we could even use PreAgent or even RLAgent if it's trained
-            action = random.sample(
-                list(
-                    retrieve_services_for_recommendation(
-                        state.search_data, ordered_services
-                    )
-                ),
-                k=K,
-            )
-
-            service_engagements = {
-                s: approx_service_engagement(
-                    user, s, engaged_services, normalized_services, index_id_map
-                )
-                for s in action
-            }
-
-            rewards_dict = {
-                s: synthesize_reward(
-                    transition_rewards_df,
-                    engagement,
-                    mode=reward_generation_mode,
-                    simple_mode_threshold=simple_reward_threshold,
-                )
-                for s, engagement in service_engagements.items()
-            }
-
-            engaged_services += _get_engaged_services(rewards_dict)
-            ordered_services += _get_ordered_services(rewards_dict)
-
-            next_state = State(
-                user=user,
-                services_history=engaged_services,
-                search_data=_get_relevant_search_data(user, ordered_services, K),
-                synthetic=True,
-            ).save()
-
-            sars = Sars(
-                state=state,
-                action=action,
-                reward=list(rewards_dict.values()),
-                next_state=next_state,
-                synthetic=True,
-            ).save()
-
-            sarses.append(sars)
-            state = next_state
-
-    return sarses, synthesized_users
-
-
 def generate_synthetic_sarses(
     service_embedder: Embedder,
     K: List[int] = 3,
@@ -232,6 +115,7 @@ def generate_synthetic_sarses(
         state = State(
             user=user,
             services_history=[],
+            elastic_services=Service.objects.distinct("id"),
             search_data=SearchData().save(),
             synthetic=True,
         ).save()
@@ -242,7 +126,7 @@ def generate_synthetic_sarses(
             # Here we could even use PreAgent or even RLAgent if it's trained
             action = random.sample(
                 list(
-                    retrieve_services_for_recommendation(
+                    retrieve_services_for_synthetic_sarses(
                         state.search_data, ordered_services
                     )
                 ),
@@ -272,6 +156,7 @@ def generate_synthetic_sarses(
             next_state = State(
                 user=user,
                 services_history=engaged_services,
+                elastic_services=Service.objects.distinct("id"),
                 search_data=SearchData().save(),
                 synthetic=True,
             ).save()
diff --git a/recommender/engines/rl/training/model_evaluation_step/model_evaluation_step.py b/recommender/engines/rl/training/model_evaluation_step/model_evaluation_step.py
index f2a67043..5131c338 100644
--- a/recommender/engines/rl/training/model_evaluation_step/model_evaluation_step.py
+++ b/recommender/engines/rl/training/model_evaluation_step/model_evaluation_step.py
@@ -123,7 +123,8 @@ def _evaluate_recommendation_time(self, actor, sarses):
         for _ in range(self.time_measurement_samples):
             start = time()
             example_user = random.choice(sarses).state.user
-            example_state = create_state(example_user, SearchData())
+            elastic_services = [service.id for service in Service.objects]
+            example_state = create_state(example_user, elastic_services, SearchData())
             self._get_recommendation(actor, example_state)
             end = time()
             recommendation_durations.append(end - start)
diff --git a/recommender/engines/rl/utils.py b/recommender/engines/rl/utils.py
index 1537dbac..481d17c2 100644
--- a/recommender/engines/rl/utils.py
+++ b/recommender/engines/rl/utils.py
@@ -101,21 +101,25 @@ def get_service_indices(index_id_map: pd.DataFrame, ids: List[int]) -> List[int]
     return indices
 
 
-def create_state(user: User, search_data: SearchData) -> State:
+def create_state(
+    user: User, elastic_services: Tuple[int], search_data: SearchData
+) -> State:
     """
     Get needed information from context and create state.
 
     Args:
         user: MongoEngine User object.
+        elastic_services: Services search context from the Marketplace
         search_data: SearchData object.
 
     Returns:
-        state: State containing information about user and search_data
+        state: State containing information about user, elastic_services and search_data
     """
 
     state = State(
         user=user,
         services_history=generate_services_history(user),
+        elastic_services=elastic_services,
         search_data=search_data,
     )
 
diff --git a/recommender/errors.py b/recommender/errors.py
index 71601b05..305d91e0 100644
--- a/recommender/errors.py
+++ b/recommender/errors.py
@@ -170,6 +170,6 @@ def message(self):
         )
 
 
-class SizeOfUsersAndSearchDataError(Exception):
+class SizeOfUsersAndElasticServicesError(Exception):
     def message(self):
-        return "Length of users and search data is not equal"
+        return "Length of users and elastic_services is not equal"
diff --git a/recommender/models/recommendation.py b/recommender/models/recommendation.py
index 8e239a90..19380364 100644
--- a/recommender/models/recommendation.py
+++ b/recommender/models/recommendation.py
@@ -24,6 +24,7 @@ class Recommendation(Document):
     panel_id = StringField()
     engine_version = StringField(blank=True)
     services = ListField(ReferenceField(Service))
+    elastic_services = ListField(ReferenceField(Service))
     search_data = ReferenceField(SearchData)
     processed = BooleanField(blank=True)
 
diff --git a/recommender/models/state.py b/recommender/models/state.py
index e3c014dc..b6cc81df 100644
--- a/recommender/models/state.py
+++ b/recommender/models/state.py
@@ -15,5 +15,6 @@
 class State(Document):
     user = ReferenceField(User, blank=True)
     services_history = ListField(ReferenceField(Service))
+    elastic_services = ListField(ReferenceField(Service))
     search_data = ReferenceField(SearchData)
     synthetic = BooleanField(default=False)
diff --git a/recommender/services/deserializer.py b/recommender/services/deserializer.py
index 0e3b6ee6..33ad4814 100644
--- a/recommender/services/deserializer.py
+++ b/recommender/services/deserializer.py
@@ -46,6 +46,7 @@ def deserialize_recommendation(cls, json_dict):
             panel_id=json_dict.get("panel_id"),
             engine_version=json_dict.get("engine_version"),
             services=json_dict.get("services"),
+            elastic_services=json_dict.get("elastic_services"),
             search_data=search_data,
         )
 
diff --git a/recommender/services/fts.py b/recommender/services/fts.py
index 94f37626..e0043f41 100644
--- a/recommender/services/fts.py
+++ b/recommender/services/fts.py
@@ -2,7 +2,7 @@
 
 """Mongo FTS Operations module"""
 
-from typing import Iterable, Optional
+from typing import Iterable, Optional, List, Tuple
 from mongoengine import QuerySet
 from recommender.models import Service, SearchData
 
@@ -44,6 +44,32 @@ def filter_services(search_data: SearchData) -> QuerySet:
 
 
 def retrieve_services_for_recommendation(
+    elastic_services: Tuple[int], accessed_services: Optional[Iterable] = None
+) -> QuerySet:
+    """
+    Selecting candidates for recommendation
+
+    Args:
+        elastic_services: Marketplace's context
+        accessed_services: Services that user accessed
+    """
+    q = list(elastic_services)
+    q = filter_available_and_recommendable_services(q)
+    q = q(id__nin=[s.id for s in accessed_services]) if accessed_services else q
+    return q
+
+
+def filter_available_and_recommendable_services(q: List[int]) -> QuerySet:
+    """
+    - Check if services exist in the RS database.
+    - Check if the status of a service allows the service to be recommended
+    """
+    q = Service.objects(id__in=q)
+    q = q(status__in=AVAILABLE_FOR_RECOMMENDATION)
+    return q
+
+
+def retrieve_services_for_synthetic_sarses(
     search_data: SearchData, accessed_services: Optional[Iterable] = None
 ):
     """Applies search info from MP and filters MongoDB by them"""
diff --git a/recommender/utils.py b/recommender/utils.py
index 444c24df..e36f1839 100644
--- a/recommender/utils.py
+++ b/recommender/utils.py
@@ -5,7 +5,6 @@
 
 import json
 import random
-import functools
 from time import time
 from datetime import datetime
 from typing import Dict, List, Union, Optional, Any
@@ -14,10 +13,9 @@
 import graphviz
 from bson import SON, ObjectId
 from mongoengine import Document
-from torch.utils.tensorboard import SummaryWriter
 from tqdm.auto import tqdm
 
-from definitions import ROOT_DIR, RUN_DIR
+from definitions import ROOT_DIR
 from recommender.engines.panel_id_to_services_number_mapping import PANEL_ID_TO_K
 from recommender.engines.rl.utils import _get_visit_ids
 from recommender.models import User, Service, UserAction, Recommendation
@@ -210,58 +208,6 @@ def load_examples() -> Dict:
     return examples
 
 
-def timeit(func):
-    if "performance_measurements" not in globals():
-        global performance_measurements
-        performance_measurements = {}
-
-    if "writer" not in globals():
-        global writer
-        writer = SummaryWriter(log_dir=RUN_DIR)
-
-    @functools.wraps(func)
-    def newfunc(*args, **kwargs):
-        start = time()
-        ret_val = func(*args, **kwargs)
-        end = time()
-        elapsed = end - start
-
-        key = f"{func.__name__}"
-        if isinstance(performance_measurements.get(key), list):
-            performance_measurements[key].append(elapsed)
-        elif performance_measurements.get(key) is None:
-            performance_measurements[key] = [elapsed]
-
-        step = len(performance_measurements[key])
-        writer.add_scalars("Performance", {key: elapsed}, step)
-        writer.flush()
-
-        return ret_val
-
-    return newfunc
-
-
-def show_times():
-    if "performance_measurements" not in globals():
-        global performance_measurements
-        performance_measurements = {}
-
-    for key, value in performance_measurements.items():
-        records = len(value)
-        mean = sum(value) / records
-        logger.info(
-            "[%s] Mean execution time: %s, records number: %s", key, mean, records
-        )
-
-    return performance_measurements
-
-
-def clear_times():
-    if "performance_measurements" not in globals():
-        global performance_measurements
-    performance_measurements = {}
-
-
 def visualize_uas(filename=None, view=True, save=False):
     """Visualize all user actions as a graph in the svg file"""
 
diff --git a/tests/endpoints/test_recommendations.py b/tests/endpoints/test_recommendations.py
index 1ce75c4d..6ff41692 100644
--- a/tests/endpoints/test_recommendations.py
+++ b/tests/endpoints/test_recommendations.py
@@ -16,6 +16,7 @@ def recommendation_data():
         "page_id": "some_page_identifier",
         "panel_id": "v1",
         "engine_version": "NCF",
+        "elastic_services": [1, 2, 3],
         "search_data": {
             "q": "Cloud GPU",
             "categories": [1],
diff --git a/tests/engine/preprocessing/test_search_data_encoder.py b/tests/engine/preprocessing/test_search_data_encoder.py
deleted file mode 100644
index 77c26b57..00000000
--- a/tests/engine/preprocessing/test_search_data_encoder.py
+++ /dev/null
@@ -1,117 +0,0 @@
-# pylint: disable-all
-
-import pytest
-import torch
-
-from recommender.engines.rl.ml_components.search_data_encoder import (
-    SearchDataEncoder,
-)
-from recommender.models import SearchData
-from tests.factories.marketplace import ServiceFactory, UserFactory
-from tests.factories.marketplace.category import CategoryFactory
-from tests.factories.marketplace.scientific_domain import ScientificDomainFactory
-
-
-@pytest.fixture
-def categories(mongo):
-    return CategoryFactory.create_batch(4)
-
-
-@pytest.fixture
-def scientific_domains(mongo):
-    return ScientificDomainFactory.create_batch(4)
-
-
-@pytest.fixture
-def services(mongo, categories, scientific_domains):
-    return [
-        ServiceFactory(
-            id=2, categories=categories[:2], scientific_domains=scientific_domains[2:]
-        ),  # i=1
-        ServiceFactory(
-            id=8, categories=categories, scientific_domains=scientific_domains
-        ),  # i=4
-        ServiceFactory(
-            id=4, categories=categories[2:], scientific_domains=[scientific_domains[3]]
-        ),  # i=2
-        ServiceFactory(
-            id=6, categories=[categories[3]], scientific_domains=scientific_domains
-        ),  # i=3
-        ServiceFactory(id=11, status="draft"),  # i=5
-        ServiceFactory(id=1, status="errored"),  # i=0
-    ]
-
-
-@pytest.fixture
-def users(mongo, services):
-    return [
-        UserFactory(accessed_services=[2, 1]),
-        UserFactory(accessed_services=[6, 11]),
-    ]
-
-
-@pytest.fixture
-def encoder(mongo, services):
-    return SearchDataEncoder()
-
-
-def test_search_data_encoder_init(encoder, services, users):
-    assert encoder.I == len(services)
-    assert encoder.forbidden_service_indices == [0, 5]
-
-
-def test_search_data_encoder_empty_search(encoder, services, users):
-    desired_mask = torch.Tensor([[0, 0, 1, 1, 1, 0], [0, 1, 1, 0, 1, 0]])
-
-    encoder_mask = encoder(users, [SearchData()] * len(users))
-    assert torch.equal(encoder_mask, desired_mask)
-
-
-def test_search_data_encoder_full_search(
-    encoder, services, users, categories, scientific_domains
-):
-    search_data = SearchData(
-        categories=categories, scientific_domains=scientific_domains
-    )
-
-    desired_mask = torch.Tensor([[0, 0, 1, 1, 1, 0], [0, 1, 1, 0, 1, 0]])
-
-    encoder_mask = encoder(users, [search_data] * len(users))
-    assert torch.equal(encoder_mask, desired_mask)
-
-
-def test_search_data_encoder_single_search(
-    encoder, services, users, categories, scientific_domains
-):
-    sd1 = SearchData(categories=categories[:2])
-    sd2 = SearchData(categories=categories[2:])
-    sd3 = SearchData(scientific_domains=scientific_domains[:2])
-    sd4 = SearchData(scientific_domains=[scientific_domains[2]])
-
-    desired_mask1 = torch.Tensor(
-        [
-            [0, 0, 0, 0, 1, 0],
-            [0, 0, 1, 0, 1, 0],
-        ]
-    )
-
-    desired_mask2 = torch.Tensor([[0, 0, 0, 1, 1, 0], [0, 1, 0, 0, 1, 0]])
-
-    assert torch.equal(encoder(users, [sd1, sd2]), desired_mask1)
-    assert torch.equal(encoder(users, [sd3, sd4]), desired_mask2)
-
-
-def test_search_data_encoder_multiple_search(
-    encoder, services, users, categories, scientific_domains
-):
-    sd1 = SearchData(
-        categories=[categories[1]], scientific_domains=[scientific_domains[2]]
-    )
-    sd2 = SearchData(
-        categories=[categories[0]], scientific_domains=[scientific_domains[3]]
-    )
-
-    desired_mask = torch.Tensor([[0, 0, 0, 0, 1, 0], [0, 1, 0, 0, 1, 0]])
-
-    encoder_mask = encoder(users, [sd1, sd2])
-    assert torch.equal(encoder_mask, desired_mask)
diff --git a/tests/engine/preprocessing/test_service_encoder.py b/tests/engine/preprocessing/test_service_encoder.py
new file mode 100644
index 00000000..f09b205f
--- /dev/null
+++ b/tests/engine/preprocessing/test_service_encoder.py
@@ -0,0 +1,132 @@
+# pylint: disable-all
+
+import pytest
+import torch
+
+from recommender.engines.rl.ml_components.service_encoder import (
+    ServiceEncoder,
+)
+from tests.factories.search_data import SearchDataFactory
+from tests.factories.state import StateFactory
+from tests.factories.marketplace import ServiceFactory, UserFactory
+from tests.factories.marketplace.category import CategoryFactory
+from tests.factories.marketplace.scientific_domain import ScientificDomainFactory
+from recommender.errors import SizeOfUsersAndElasticServicesError
+
+
+@pytest.fixture
+def categories(mongo):
+    return CategoryFactory.create_batch(4)
+
+
+@pytest.fixture
+def scientific_domains(mongo):
+    return ScientificDomainFactory.create_batch(4)
+
+
+@pytest.fixture
+def services(mongo, categories, scientific_domains):
+    return [
+        ServiceFactory(
+            id=2, categories=categories[:2], scientific_domains=scientific_domains[2:]
+        ),  # i=1
+        ServiceFactory(
+            id=8, categories=categories, scientific_domains=scientific_domains
+        ),  # i=4
+        ServiceFactory(
+            id=4, categories=categories[2:], scientific_domains=[scientific_domains[3]]
+        ),  # i=2
+        ServiceFactory(
+            id=6, categories=[categories[3]], scientific_domains=scientific_domains
+        ),  # i=3
+        ServiceFactory(id=11, status="draft"),  # i=5
+        ServiceFactory(id=1, status="errored"),  # i=0
+    ]
+
+
+@pytest.fixture
+def states(services):
+    return [
+        StateFactory(
+            elastic_services=[services[0], services[1], services[-1]],
+            search_data=SearchDataFactory(q=None),
+            non_empty_history=True,
+        ),
+        StateFactory(
+            elastic_services=[services[0], services[1], services[-1]],
+            search_data=SearchDataFactory(q=None),
+            non_empty_history=True,
+        ),
+    ]
+
+
+@pytest.fixture
+def users(mongo, services):
+    return [
+        UserFactory(accessed_services=[2, 1]),
+        UserFactory(accessed_services=[6, 11]),
+    ]
+
+
+@pytest.fixture
+def encoder(services):
+    return ServiceEncoder()
+
+
+def test_service_encoder_init(encoder, services, users):
+    assert encoder.I == len(services)
+    assert encoder.forbidden_service_indices == [0, 5]
+
+
+def test_service_encoder_one_state_one_user(encoder):
+    state = StateFactory(
+        elastic_services=[2, 8, 4, 6, 11, 1],
+        search_data=SearchDataFactory(q=None),
+        non_empty_history=True,
+    )
+    user = UserFactory(accessed_services=[2, 1])
+
+    desired_mask = torch.Tensor([[0, 0, 1, 1, 1, 0]])
+    encoder_mask = encoder([user], [state])
+
+    assert torch.equal(encoder_mask, desired_mask)
+
+
+def test_service_encoder_multiple_states_users(encoder, states, users):
+    desired_mask = torch.Tensor([[0, 0, 0, 0, 1, 0], [0, 1, 0, 0, 1, 0]])
+
+    assert torch.equal(encoder(users, states), desired_mask)
+
+
+def test_service_encoder_exception(encoder):
+    state = StateFactory(
+        elastic_services=[2, 8, 4, 6, 11, 1],
+        search_data=SearchDataFactory(q=None),
+        non_empty_history=True,
+    )
+    user = [
+        UserFactory(accessed_services=[2, 1]),
+        UserFactory(accessed_services=[8, 2]),
+    ]
+
+    # 1 state and 2 users
+    with pytest.raises(SizeOfUsersAndElasticServicesError):
+        encoder(user, state)
+
+    state = [
+        StateFactory(
+            elastic_services=[2, 8, 4, 6, 11, 1],
+            search_data=SearchDataFactory(q=None),
+            non_empty_history=True,
+        ),
+        StateFactory(
+            elastic_services=[2, 8, 1],
+            search_data=SearchDataFactory(q=None),
+            non_empty_history=True,
+        ),
+    ]
+    user = [UserFactory(accessed_services=[2, 1])]
+
+    # 2 states and 1 user
+    with pytest.raises(SizeOfUsersAndElasticServicesError):
+        encoder(user, state)
diff --git a/tests/engines/autoencoders/ml_components/test_normalizer.py b/tests/engines/autoencoders/ml_components/test_normalizer.py
index d506d7f3..7c675069 100644
--- a/tests/engines/autoencoders/ml_components/test_normalizer.py
+++ b/tests/engines/autoencoders/ml_components/test_normalizer.py
@@ -83,7 +83,6 @@ def test_normalizer_call_on_tensors_dim_wise(batch):
 def test_normalizer_call_on_tensors_norm_wise(batch):
     normalizer = Normalizer(mode=NormalizationMode.NORM_WISE)
     output, _ = normalizer(batch)
-    print(output)
     desired = torch.Tensor(
         [
             [0.2411, -0.1002, -0.2520],
diff --git a/tests/engines/ncf/inference/test_ncf_inference_component.py b/tests/engines/ncf/inference/test_ncf_inference_component.py
index 5d3a5070..802fd9ae 100644
--- a/tests/engines/ncf/inference/test_ncf_inference_component.py
+++ b/tests/engines/ncf/inference/test_ncf_inference_component.py
@@ -1,6 +1,7 @@
 # pylint: disable-all
 
 import pytest
+import random
 from torch import Tensor
 
 from recommender.engines.autoencoders.ml_components.embedder import (
@@ -39,16 +40,26 @@ def test_ncf_inference_component(
 
     with pytest.raises(NoSavedMLComponentError):
         K = 2
-        ncf_inference_component = NCFInferenceComponent(K)
+        NCFInferenceComponent(K)
+
+    elastic_services = [service.id for service in Service.objects]
 
     # With-model case
     ncf_model.save(version=NEURAL_CF)
+
     for panel_id_version, K in list(PANEL_ID_TO_K.items()):
         ncf_inference_component = NCFInferenceComponent(K)
-        user = User.objects.first()
-        context = {"panel_id": panel_id_version, "search_data": {}, "user_id": user.id}
+        user = random.choice(list(User.objects))
+
+        context = {
+            "panel_id": panel_id_version,
+            "elastic_services": elastic_services,
+            "search_data": {},
+            "user_id": user.id,
+        }
 
         services_ids_1 = ncf_inference_component(context)
+
         assert isinstance(
             ncf_inference_component.neural_cf_model, NeuralCollaborativeFilteringModel
         )
@@ -56,13 +67,20 @@ def test_ncf_inference_component(
         assert isinstance(services_ids_1, list)
         assert len(services_ids_1) == PANEL_ID_TO_K.get(context["panel_id"])
         assert all([isinstance(service_id, int) for service_id in services_ids_1])
+        assert all(service in elastic_services for service in services_ids_1)
 
         services_ids_2 = ncf_inference_component(context)
         assert services_ids_1 == services_ids_2
 
     for panel_id_version, K in list(PANEL_ID_TO_K.items()):
         ncf_inference_component = NCFInferenceComponent(K)
-        context = {"panel_id": panel_id_version, "search_data": {}, "user_id": -1}
+
+        context = {
+            "panel_id": panel_id_version,
+            "elastic_services": elastic_services,
+            "search_data": {},
+            "user_id": -1,
+        }
 
         services_ids_1 = ncf_inference_component(context)
         assert isinstance(
@@ -72,13 +90,19 @@ def test_ncf_inference_component(
         assert isinstance(services_ids_1, list)
         assert len(services_ids_1) == PANEL_ID_TO_K.get(context["panel_id"])
         assert all([isinstance(service_id, int) for service_id in services_ids_1])
+        assert all(service in elastic_services for service in services_ids_1)
 
         services_ids_2 = ncf_inference_component(context)
         assert services_ids_1 != services_ids_2
 
     for panel_id_version, K in list(PANEL_ID_TO_K.items()):
         ncf_inference_component = NCFInferenceComponent(K)
-        context = {"panel_id": panel_id_version, "search_data": {}}
+
+        context = {
+            "panel_id": panel_id_version,
+            "elastic_services": elastic_services,
+            "search_data": {},
+        }
 
         services_ids_1 = ncf_inference_component(context)
         assert isinstance(
@@ -88,6 +112,7 @@ def test_ncf_inference_component(
         assert isinstance(services_ids_1, list)
         assert len(services_ids_1) == PANEL_ID_TO_K.get(context["panel_id"])
         assert all([isinstance(service_id, int) for service_id in services_ids_1])
+        assert all(service in elastic_services for service in services_ids_1)
 
         services_ids_2 = ncf_inference_component(context)
         assert services_ids_1 != services_ids_2
diff --git a/tests/engines/rl/inference/test_rl_inference_component.py b/tests/engines/rl/inference/test_rl_inference_component.py
index 0e1330f1..36ac8479 100644
--- a/tests/engines/rl/inference/test_rl_inference_component.py
+++ b/tests/engines/rl/inference/test_rl_inference_component.py
@@ -6,7 +6,7 @@
 from recommender import User
 from recommender.engines.panel_id_to_services_number_mapping import PANEL_ID_TO_K
 from recommender.engines.rl.inference.rl_inference_component import RLInferenceComponent
-from recommender.models import Category
+from recommender.models import Category, Service
 
 
 @pytest.mark.parametrize("ver", ["v1", "v2"])
@@ -15,13 +15,21 @@ def test_known_user(mongo, generate_users_and_services, mock_rl_pipeline_exec, v
     inference_component = RLInferenceComponent(K=K)
 
     user = random.choice(list(User.objects))
-    context = {"panel_id": ver, "search_data": {}, "user_id": user.id}
+    elastic_services = [service.id for service in Service.objects]
+
+    context = {
+        "panel_id": ver,
+        "elastic_services": elastic_services,
+        "search_data": {},
+        "user_id": user.id,
+    }
 
     services_ids_1 = inference_component(context)
 
     assert isinstance(services_ids_1, list)
     assert len(services_ids_1) == PANEL_ID_TO_K.get(context["panel_id"])
     assert all([isinstance(service_id, int) for service_id in services_ids_1])
+    assert all(service in elastic_services for service in services_ids_1)
 
     services_ids_2 = inference_component(context)
     assert services_ids_1 == services_ids_2
@@ -31,10 +39,20 @@ def test_known_user(mongo, generate_users_and_services, mock_rl_pipeline_exec, v
 def test_unknown_user(mongo, generate_users_and_services, mock_rl_pipeline_exec, ver):
     K = PANEL_ID_TO_K[ver]
     inference_component = RLInferenceComponent(K=K)
+    elastic_services = [service.id for service in Service.objects]
 
     contexts = [
-        {"panel_id": ver, "search_data": {}, "user_id": -1},
-        {"panel_id": ver, "search_data": {}},
+        {
+            "panel_id": ver,
+            "elastic_services": elastic_services,
+            "search_data": {},
+            "user_id": -1,
+        },
+        {
+            "panel_id": ver,
+            "elastic_services": elastic_services,
+            "search_data": {},
+        },
     ]
 
     for context in contexts:
@@ -46,28 +64,6 @@ def test_unknown_user(mongo, generate_users_and_services, mock_rl_pipeline_exec,
         assert services_ids_1 != services_ids_2
 
 
-@pytest.mark.parametrize("ver", ["v1", "v2"])
-def test_not_empty_search_data(
-    mongo, generate_users_and_services, mock_rl_pipeline_exec, ver
-):
-    K = PANEL_ID_TO_K[ver]
-    inference_component = RLInferenceComponent(K=K)
-
-    user = User.objects.first()
-    context = {
-        "panel_id": ver,
-        "search_data": {"categories": list(Category.objects.distinct("id"))},
-        "user_id": user.id,
-    }
-
-    services_ids_1 = inference_component(context)
-    _check_proper(services_ids_1, ver)
-
-    services_ids_2 = inference_component(context)
-    _check_proper(services_ids_2, ver)
-    assert services_ids_1 == services_ids_2
-
-
 def _check_proper(services, ver):
     assert isinstance(services, list)
     assert len(services) == PANEL_ID_TO_K.get(ver)
diff --git a/tests/services/synthetic_dataset/test_dataset.py b/tests/services/synthetic_dataset/test_dataset.py
index e13548db..f315cd8c 100644
--- a/tests/services/synthetic_dataset/test_dataset.py
+++ b/tests/services/synthetic_dataset/test_dataset.py
@@ -38,10 +38,3 @@ def test__normalize_embedded_services(service_embeddings, normalized_embedded_se
         normalized_embedded_services,
         atol=10e-5,
     )
-
-
-@pytest.mark.skip(
-    reason="no way of currently testing this as it requires infeasible amount of setup"
-)
-def test_generate_dataset():
-    pass
diff --git a/tests/services/test_fts.py b/tests/services/test_fts.py
index 34c6eaf9..8c7f913b 100644
--- a/tests/services/test_fts.py
+++ b/tests/services/test_fts.py
@@ -1,8 +1,10 @@
 # pylint: disable-all
+import pytest
 
 from recommender.models import Service, SearchData
 from recommender.services.fts import (
     retrieve_services_for_recommendation,
+    retrieve_services_for_synthetic_sarses,
     retrieve_forbidden_services,
     filter_services,
 )
@@ -14,6 +16,19 @@
 from tests.factories.marketplace.target_user import TargetUserFactory
 
 
+@pytest.fixture
+def create_services(mongo):
+    services = [
+        ServiceFactory(status="published"),
+        ServiceFactory(status="unverified"),
+        ServiceFactory(status="unverified"),
+        ServiceFactory(status="errored"),
+        ServiceFactory(status="deleted"),
+        ServiceFactory(status="draft"),
+    ]
+    return services
+
+
 def test_filter_services(mongo, mocker):
     categories = CategoryFactory.create_batch(2)
     providers = ProviderFactory.create_batch(2)
@@ -102,32 +117,32 @@ def test_filter_services(mongo, mocker):
     ) == [s1]
 
 
-def test_retrieve_forbidden_services(mongo):
-    services = [
-        ServiceFactory(status="published"),
-        ServiceFactory(status="unverified"),
-        ServiceFactory(status="errored"),
-        ServiceFactory(status="deleted"),
-        ServiceFactory(status="draft"),
-        ServiceFactory(status="unverified"),
-    ]
+def test_retrieve_forbidden_services(create_services):
+    services = create_services
 
-    assert list(retrieve_forbidden_services()) == services[2:5]
+    assert list(retrieve_forbidden_services()) == services[3:6]
 
 
-def test_retrieve_services_for_recommendations(mongo):
-    services = [
-        ServiceFactory(status="published"),
-        ServiceFactory(status="published"),
-        ServiceFactory(status="unverified"),
-        ServiceFactory(status="unverified"),
-        ServiceFactory(status="deleted"),
-        ServiceFactory(status="draft"),
-    ]
+def test_retrieve_services_for_recommendation(create_services):
+    services = create_services
+    services_id = []
 
-    assert list(retrieve_services_for_recommendation(SearchData())) == services[:4]
+    [services_id.append(service.id) for service in services]
+
+    assert list(retrieve_services_for_recommendation(services_id)) == services[:3]
     assert list(
         retrieve_services_for_recommendation(
+            services_id, accessed_services=[services[0], services[2]]
+        )
+    ) == [services[1]]
+
+
+def test_retrieve_services_for_synthetic_sarses(create_services):
+    services = create_services
+
+    assert list(retrieve_services_for_synthetic_sarses(SearchData())) == services[:3]
+    assert list(
+        retrieve_services_for_synthetic_sarses(
             SearchData(), accessed_services=[services[0], services[2]]
         )
-    ) == [services[1], services[3]]
+    ) == [services[1]]