Skip to main content

Similarity / Vector search

LICENSE Weaviate on Stackoverflow badge Weaviate issues on GitHub badge Weaviate version badge Weaviate total Docker pulls badge Go Report Card

Overview​

This page shows you how to perform similarity-based searches with the nearXXX operators.

These operators work by searching for objects with the most similar vector representation to the query. Note that due to differences in configuration and data, we have a separate page for image searches.

Related pages

Similarity-based operators​

These operators are available:

  • near<Media> - Finds objects closest to an input medium:
    • E.g.: Use it to find text objects most similar to cute animals, or images most similar to a particular image.
  • nearObject - Finds objects closest to another Weaviate object:
    • E.g.: Use it to find Weaviate objects most similar to object 56b9449e-65db-5df4-887b-0a4773f52aa7.
  • nearVector - Find objects closest to an input vector.
    • E.g.: Use it to find Weaviate objects most similar to vector [-0.368, 0.1397, ... , 0.0971].

An input medium​

This is only available for classes with a vectorizer configured.

You can use these operators to find objects most similar to a raw (un-vectorized) input, such as text or image. For text objects, you can provide an input text to nearText, and for image objects, you can provide an input image to nearImage. (Or either if you are using CLIP.)

The example below searches the JeopardyQuestion class for the top 2 objects best matching "animals in movies", using nearText:

response = (
    client.query
    .get("JeopardyQuestion", ["question", "answer"])
    .with_near_text({
        "concepts": ["animals in movies"]
    })
    .with_limit(2)
    .with_additional(["distance"])
    .do()
)

print(json.dumps(response, indent=2))
Example response

It should produce a response like the one below:

{
  "data": {
    "Get": {
      "JeopardyQuestion": [
        {
          "answer": "meerkats",
          "question": "Group of mammals seen <a href=\"http://www.j-archive.com/media/1998-06-01_J_28.jpg\" target=\"_blank\">here</a>:  [like Timon in <i>The Lion King</i>]",
          "_additional": { "distance": 0.17602634 }
        },
        {
          "answer": "dogs",
          "question": "Scooby-Doo, Goofy & Pluto are cartoon versions",
          "_additional": { "distance": 0.17842108 }
        }
      ]
    }
  }
}

An object​

You can use the nearObject operator to find objects most similar to an existing Weaviate object. To do so, specify the object ID (e.g. 56b9449e-65db-5df4-887b-0a4773f52aa7) as shown below.

How to retrieve object IDs

See this section

The example below searches the JeopardyQuestion class for the top 2 objects best matching the object with ID 56b9449e-65db-5df4-887b-0a4773f52aa7, using nearObject:

response = (
    client.query
    .get("JeopardyQuestion", ["question", "answer"])
    .with_near_object({
        "id": "56b9449e-65db-5df4-887b-0a4773f52aa7"
    })
    .with_limit(2)
    .with_additional(["distance"])
    .do()
)

print(json.dumps(response, indent=2))

A vector​

You can use the nearVector operator to find objects most similar to an input vector (e.g. [-0.368, 0.1397, ... , 0.0971]).

The example below searches the JeopardyQuestion class for the top 2 objects best matching the object with the provided vector, using nearVector:

response = (
    client.query
    .get("JeopardyQuestion", ["question", "answer"])
    .with_near_vector({
        "vector": [-0.0125526935, -0.021168863, -0.01076519, -0.02589537, -0.0070362035, 0.019870078, -0.010001986, -0.019120263, 0.00090044655, -0.017393013, 0.021302758, 0.010055545, 0.02937665, -0.003816019, 0.007692291, 0.012385325, 0.032750815, 0.020847514, 0.020311933, -0.022159688, -0.0009924996, 0.009399457, 0.0022226637, -0.029510546, 0.014393755, -0.007223657, 0.018276723, -0.03639277, -0.010001986, -0.022842556, 0.010363504, -0.020927852, -0.006929087, -0.022521207, -0.007652122, -0.011126708, 0.0279038, -0.01721895, 0.016482525, 0.002281243, -0.00169294, 0.009191919, -0.019655844, -0.022869334, -0.012412104, 0.0031967526, -0.0033457114, -0.01483561, -0.03173321, 0.004746592, 0.010095714, 0.007973471, -0.032134898, -0.023739655, -0.008040419, 0.018290112, -0.013637247, -0.008488968, 0.024623364, -0.039365247, -0.0032586793, 0.0009606995, -0.029510546, 0.0063265576, -0.019602288, 0.003081268, 0.013463182, -0.006601043, 0.019910246, -0.01542475, 0.0367409, -0.01193008, 0.012961075, -0.015625594, 0.0062462203, -0.0058646183, -0.0059248717, 0.01889264, 0.008127451, 0.0037155973, 0.037142586, -0.025373178, -0.005503101, 0.014982895, 0.035053816, -0.012432188, -0.017285896, 0.022936283, 0.0024620018, 0.016937768, -0.0062127467, 0.02154377, 0.0066378643, 0.029698, 0.0013071538, 0.0043850746, -0.008040419, 0.024797428, -0.012452273, -0.025132166, -0.0031900578, 0.0000019433794, -0.002378317, -0.008629559, 0.0126732, -0.0022494427, 0.0009623732, 0.0035582704, 0.017312676, -0.024569806, -0.008890655, 0.023056788, 0.014902558, -0.047104403, -0.009011161, -0.030447815, 0.017982153, -0.0042009684, -0.00654079, 0.00069249026, 0.011936775, 0.023378137, 0.025105387, -0.009245478, 0.030929837, 0.00394322, 0.02123581, -0.0042545265, 0.0022578111, -0.017259117, 0.047157962, -0.00022029977, 0.03497348, -0.00072094303, -0.023605758, 0.036499888, -0.015384582, 0.011099929, -0.0139519, -0.03408977, 0.013155223, 0.030501373, -0.026698742, 0.004311432, -0.010236303, 0.011361024, 0.023793213, -0.00014874942, 0.0020352101, 0.0026829292, 0.00989487, 0.0074780583, 0.02734144, 0.003826061, 0.011722542, 0.00712993, -0.013992069, 0.0009406152, 0.010785274, -0.012325072, 0.01692438, 0.010617905, 0.016750315, -0.0070295087, 0.017687583, 0.038320865, 0.020485997, 0.005054551, -0.018812304, 0.0007201062, 0.0015381235, 0.0349467, 0.014728494, 0.050773136, -0.017901815, 0.0027716348, 0.0064704954, 0.026671965, -0.015063233, -0.013536825, 0.016696757, 0.008127451, 0.026966535, 0.029912233, -0.0031431946, 0.015156959, 0.012412104, -0.047907773, 0.022012403, -0.027006702, -0.0069491714, 0.010718327, 0.011976943, -0.008127451, -0.65212417, 0.00024289463, 0.0051214993, -0.013007938, 0.022373922, 0.0337952, -0.0026829292, -0.0110463705, -0.013034717, -0.0012167745, 0.010062239, -0.0023013272, 0.024409132, -0.009118277, -0.020191427, -0.01597372, 0.010115798, -0.030929837, -0.010932559, 0.010912475, -0.0009841312, 0.010571042, -0.008348378, -0.009104887, 0.02711382, 0.0036553445, -0.018263333, -0.030876279, 0.014594599, 0.037704945, -0.030126465, 0.014366977, 0.0055533117, 0.003487975, 0.044988856, 0.009881481, -0.012699978, 0.041132666, 0.01744657, 0.05417408, -0.004686339, 0.016121006, 0.0070495927, 0.015478308, -0.020593112, 0.0012376956, 0.027127208, -0.0051248465, 0.0005979267, 0.0063366, -0.008616169, 0.027877023, -0.00042679158, 0.008442105, 0.00069751136, 0.023806602, 0.029296314, -0.0047332025, 0.027877023, 0.0033005215, 0.014996285, -0.0061424514, 0.00451897, 0.015531867, -0.015317634, 0.044185482, 0.010196134, 0.007504837, 0.012405409, -0.030126465, 0.03821375, 0.0256008, -0.016710145, 0.0032804373, -0.013884953, 0.022775607, 0.030608488, -0.023431696, -0.008502358, 0.008683117, -0.0045490963, -0.0030143203, -0.024074392, 0.00874337, 0.009466405, -0.0072370465, -0.021383096, 0.001360712, 0.020298542, 0.0040168623, 0.008201093, 0.011106623, -0.03202778, 0.0046461704, -0.00088370964, -0.008957602, 0.0057575023, 0.00037407028, 0.017259117, -0.0482559, -0.0049507823, -0.024235068, -0.0014418861, 0.004425243, 0.023244241, 0.0107919695, -0.017058274, 0.0183035, 0.033339955, -0.009091497, 0.000118936776, 0.0031900578, -0.000044483608, -0.017058274, 0.001529755, -0.027984139, 0.02740839, -0.015344413, 0.015264076, -0.01719217, 0.010463926, -0.0067048124, 0.014942727, -0.00026653553, 0.02677908, -0.00036570182, -0.043194655, -0.022855945, -0.011294077, 0.005764197, 0.004910614, -0.0029724778, 0.0056637754, -0.01425986, -0.000008708432, 0.01866502, 0.031626094, 0.0050378144, 0.015451529, 0.009406152, -0.030742384, -0.0024318753, -0.029751558, -0.008348378, 0.0028519721, -0.008388547, -0.010611211, 0.0139519, -0.0006895613, -0.001230164, -0.0062462203, -0.013510046, 0.010617905, -0.010229609, 0.022213247, -0.00610563, -0.00568386, -0.0056503857, 0.02416812, -0.0076253433, 0.015183738, -0.005188447, -0.016080838, 0.013516741, 0.0062897364, -0.0068520973, 0.021396484, 0.007799407, -0.01721895, -0.025266062, 0.013791226, -0.017205559, -0.002068684, 0.032938268, 0.014661547, 0.023552202, -0.005827797, -0.008442105, -0.0074914475, 0.009111582, 0.016817262, -0.0050244248, -0.005871313, -0.008368462, 0.040329296, 0.008683117, 0.031518977, 0.026109602, -0.025815032, 0.011006202, -0.0034310697, 0.019575508, -0.013831395, -0.008676422, -0.008770149, -0.019990584, 0.008750064, 0.02851972, 0.0337952, 0.012666505, 0.021383096, -0.027448557, 0.0035448808, -0.016214734, 0.015197128, -0.027582452, -0.0138046155, -0.03899034, 0.008261346, 0.015478308, 0.017888425, 0.0153979715, 0.010658074, -0.011581952, 0.02530623, 0.017982153, -0.0059449556, 0.0054294583, 0.0022879376, -0.018758746, -0.0076119537, -0.027689569, 0.013463182, 0.011186961, -0.0063165156, 0.028412605, 0.011347636, 0.008709895, -0.003374164, -0.007919913, -0.025828423, 0.0033875536, -0.013831395, -0.0035716598, 0.010450536, -0.025172336, 0.003990083, -0.00093224674, 0.024047613, 0.008027029, -0.0029440252, 0.023458473, 0.016643198, -0.0326437, 0.019147042, 0.01925416, -0.0020151257, 0.0038628823, -0.026738912, 0.0008753412, -0.025105387, 0.0069491714, -0.02623011, 0.027033482, -0.0040737675, -0.021034967, 0.019468391, 0.0026042655, 0.03467891, 0.016107617, -0.0057139862, -0.011735932, 0.017687583, 0.011628816, 0.015090012, -0.006678033, -0.011715848, -0.01833028, 0.008040419, -0.01921399, -0.03267048, -0.005914829, 0.0014435598, -0.0030662047, 0.005479669, 0.01597372, -0.01454104, 0.023257632, 0.019722793, 0.0344379, 0.006929087, -0.043248214, 0.015853215, 0.012766927, -0.007417805, -0.018316891, -0.01163551, -0.017352844, -0.01978974, 0.015304244, -0.00005920687, 0.033580966, -0.0022343795, 0.0047800657, -0.007357552, 0.00033536615, 0.00887057, -0.025654359, 0.016388796, -0.011361024, 0.00019090556, 0.0060119033, -0.010075629, -0.0131485285, 0.01604067, -0.015531867, 0.0035616176, -0.017259117, 0.0035415334, 0.009265562, -0.0043348637, -0.005867966, -0.03283115, -0.004773371, -0.018410617, -0.0095400475, -0.006520706, -0.00414741, 0.031197628, 0.013690805, -0.008984381, -0.022320364, -0.012492441, -0.005724028, 0.09806499, 0.017272506, -0.00007704216, 0.00858939, 0.0030126465, -0.002835235, -0.023753043, -0.025587412, 0.016067449, 0.0024536331, 0.004719813, -0.02908208, 0.027743127, 0.0023414958, 0.0152908545, 0.00552988, -0.031974223, 0.0019582203, 0.010812053, -0.01952195, -0.00006171741, -0.02241409, 0.025252672, 0.013737668, 0.002356559, -0.03719614, 0.021637497, 0.033580966, 0.0044453274, -0.0074378895, -0.014715104, -0.01741979, -0.013489962, -0.003221858, 0.0038561875, -0.013121749, -0.012974464, 0.012619642, 0.053424265, -0.020459218, 0.011581952, 0.041962817, -0.00087032013, -0.0036988605, -0.0010025419, -0.020392269, 0.014902558, 0.021409875, 0.01771436, -0.006483885, 0.036633782, -0.00028808432, 0.011983639, 0.014326808, 0.024931323, 0.002629371, -0.01223804, -0.010972728, -0.011253908, 0.013831395, -0.01748674, -0.013777837, -0.0043449057, -0.009292341, -0.0015849868, -0.019455003, -0.031170849, -0.014393755, -0.03778528, -0.0028335615, -0.00785966, -0.027528895, -0.021008188, -0.03786562, -0.0008226199, -0.005539922, 0.011970249, -0.016937768, -0.0044553694, 0.015839826, -0.014929337, -0.011166876, 0.0031448682, -0.032402687, -0.011207045, -0.009432931, 0.0034059642, -0.00089124124, -0.009439626, -0.012840569, 0.013610467, 0.008877265, 0.006108978, 0.0021289368, 0.039124236, 0.0025557284, -0.004277958, 0.02822515, 0.022373922, -0.00888396, 0.032777593, -0.021610718, -0.010490704, -0.0017222296, -0.011113319, -0.024569806, 0.0024703701, 0.021155473, -0.004555791, -0.0060353354, 0.008241262, -0.03234913, -0.00048076818, -0.0069960346, 0.02910886, 0.013315897, -0.014728494, 0.01454104, -0.00567047, -0.0012602905, 0.0001736456, 0.005302258, -0.0000424961, 0.035589397, -0.01570593, 0.0107919695, 0.0051348885, -0.015331023, -0.0034193539, 0.003625218, -0.010477315, 0.024583196, -0.0030226887, -0.011776101, -0.040115062, -0.009091497, -0.003886314, 0.017888425, -0.03143864, -0.008629559, -0.005533227, -0.017138612, 0.01338954, -0.02681925, -0.006688075, -0.026538068, 0.0050210776, 0.011401193, 0.0076655117, 0.008576, -0.028171593, -0.0022025793, 0.005911482, 0.017205559, -0.02066006, -0.0413469, -0.016910989, 0.0097944485, 0.020807344, 0.030742384, 0.026738912, -0.011628816, 0.03350063, 0.011146792, -0.024556417, 0.019709403, -0.00712993, 0.012110839, -0.044694286, 0.02795736, 0.016777094, -0.0054729744, 0.025975708, 0.0109191695, 0.009821228, 0.012485746, 0.01571932, 0.0018661672, -0.014567819, -0.010972728, 0.0022394005, 0.01626829, 0.0014820547, -0.0030026045, 0.004120631, -0.023699487, 0.040918436, 0.0011640531, -0.0092856465, -0.0180491, 0.03459857, -0.013161918, -0.0036151758, -0.0073910262, 0.0028737301, -0.017968763, -0.016549472, -0.01355691, 0.0031616052, 0.0067516756, 0.0023096956, -0.0076789013, -0.009955123, 0.011233824, -0.0072906045, 0.016402187, 0.009727501, -0.0153979715, 0.020445827, -0.0042980425, -0.024556417, -0.048496913, -0.026886197, -0.047693543, 0.0007615301, -0.013925122, -0.010437147, 0.01483561, -0.0050277724, -0.022266805, 0.02793058, -0.015264076, 0.032563362, 0.00472316, 0.017526908, 0.021061746, -0.013818005, -0.021945456, 0.028573278, -0.0313583, 0.016469134, 0.00013180329, -0.000116426236, -0.0018477566, -0.03722292, -0.002868709, 0.001186648, -0.037463933, -0.046568822, 0.0128004, 0.015197128, 0.013054801, -0.017821478, -0.022320364, -0.022012403, 0.013289118, -0.0043516005, -0.0029808464, -0.01660303, -0.03786562, -0.024877766, -0.013356066, -0.006825318, 0.027582452, -0.0042545265, -0.0017063295, 0.024891155, -0.0049240035, -0.014500872, -0.016803874, 0.008127451, 0.022855945, -0.0014284966, -0.006339947, 0.01604067, 0.0026092867, 0.012057281, -0.008569306, 0.00007374708, 0.02766279, -0.025774864, 0.0047064233, -0.024676923, 0.013938512, -0.002286264, -0.011166876, -0.024074392, -0.018450785, -0.0049842563, 0.0035080595, 0.028305488, 0.033286396, -0.003054489, -0.003272069, -0.024502859, 0.021302758, -0.015558646, -0.006798539, 0.005667123, -0.01716539, 0.003325627, 0.00885718, -0.0047767186, -0.0073843314, -0.0038193662, -0.009352594, 0.0209948, 0.041507576, -0.036526665, -0.0022661798, -0.035401944, 0.012204566, -0.034759246, -0.008850486, -0.0009975208, -0.00022176426, -0.008629559, -0.015357803, -0.01455443, -0.0059416085, -0.01687082, 0.014487483, -0.0008845465, -0.0010284841, 0.02708704, 0.028653616, 0.0033189321, -0.025373178, 0.0036620393, 0.018772135, -0.0031130682, 0.0070495927, -0.00006830758, -0.017674193, 0.000969068, -0.018290112, -0.005546617, 0.0037658082, -0.00016872912, -0.024784038, -0.020860903, 0.02070023, 0.0029138986, -0.036285654, -0.041159447, -0.022106132, -0.018651629, 0.03435756, -0.008194398, -0.020485997, 0.01660303, 0.026270278, 0.0079065235, 0.0015649025, -0.005807713, -0.012733453, -0.0042377897, -0.021891898, -0.0180491, -0.008783538, -0.017111832, 0.005493059, 0.011501615, -0.0025657706, -0.018946199, 0.006052072, -0.0120438915, 0.010644685, -0.005165015, 0.009881481, 0.02677908, -0.0035716598, 0.005449543, 0.021758003, -0.0072035724, 0.010745106, -0.012130924, -0.0011799532, 0.0036620393, -0.0034411119, 0.013028023, 0.045095973, -0.021396484, -0.01895959, 0.016281681, 0.0020050837, 0.008214483, 0.004632781, -0.030501373, -0.019709403, -0.021075137, -0.0027230978, -0.015183738, 0.0008828728, 0.015304244, -0.0034578487, -0.02940343, 0.015344413, 0.00785966, -0.0026260235, -0.008529137, 0.00442859, 0.0013900016, 0.0001500047, -0.024368962, -0.005580091, -0.017205559, -0.0285465, 0.0054729744, -0.0009422889, -0.0076722065, 0.02475726, -0.02241409, -0.016469134, -0.0064370213, 0.00018034037, 0.009044634, -0.0044486746, 0.000060462142, -0.014942727, 0.026658574, -0.0043181265, 0.030046128, -0.042043157, 0.016616419, -0.007170099, 0.02040566, -0.008227873, 0.025975708, -0.027877023, -0.022668492, 0.0051181517, -0.007116541, 0.016522693, -0.0025373178, -0.0018259985, -0.015906774, 0.013858174, -0.019843299, 0.0029942358, -0.01632185, -0.029831896, -0.024007445, -0.0045022327, -0.015946941, 0.030662047, 0.18091947, -0.016576251, 0.003936525, 0.039659817, -0.008160925, 0.021168863, 0.026002487, -0.0043248213, -0.008488968, 0.0125526935, -0.007839575, 0.024020836, -0.014500872, 0.008529137, 0.0011925059, -0.015652372, -0.00050880254, -0.0032017739, -0.006353337, -0.03438434, -0.013208781, -0.0023113694, -0.011608731, -0.015411361, 0.022842556, 0.0013423014, -0.0017356192, -0.005104762, 0.0062395255, 0.0056403438, 0.0061960095, -0.033018608, 0.0053591635, -0.02067345, -0.001453602, -0.013289118, -0.02851972, 0.028118035, 0.0052687842, 0.01338954, -0.0035314912, 0.009673943, 0.009191919, 0.01281379, -0.013992069, 0.008134145, -0.004575875, 0.0015013022, -0.00028620142, 0.03550906, -0.0512016, 0.010477315, 0.008897349, 0.03347385, -0.02471709, 0.0011297425, 0.005851229, -0.019588897, 0.012037196, 0.010182745, 0.0065776114, 0.030233582, -0.01309497, 0.018839084, -0.024623364, 0.0072370465, -0.02241409, 0.03400943, -0.00069207186, -0.014674936, -0.0031833632, -0.024784038, -0.02645773, -0.012793706, -0.0008506542, -0.03583041, -0.012325072, 0.026966535, 0.01018944, 0.013356066, -0.02474387, -0.014326808, -0.007658817, 0.012827179, -0.02740839, -0.015277465, 0.021784782, -0.0015858236, -0.0018460829, 0.0004573365, -0.0057072914, -0.019588897, 0.0058411867, -0.002308022, -0.00066278223, 0.006460453, -0.00038369402, 0.018705187, -0.009078108, -0.020298542, -0.035991084, -0.047211517, 0.018571293, -0.0041775363, -0.008676422, -0.002138979, -0.007504837, -0.00078579865, 0.014621378, -0.0043850746, -0.01455443, -0.015906774, 0.0010176051, 0.006935782, 0.025199115, -0.0038093242, 0.013690805, -0.022253416, 0.036874793, -0.019053316, -0.0044821487, 0.0042377897, 0.005998514, 0.0064102425, 0.008080588, -0.028064476, -0.025239283, 0.0070295087, 0.023083568, -0.028653616, -0.010771885, -0.019280938, -0.005563354, -0.012579473, -0.005258742, 0.0012109166, 0.015531867, -0.017339455, 0.016241511, 0.0069424766, 0.015652372, 0.014380367, 0.006791844, -0.0023967277, 0.037945956, -0.0285465, 0.02128937, 0.0049942983, -0.029831896, -0.023819992, -0.016281681, -0.0031850368, 0.0029691304, -0.0038227136, 0.023645928, -0.036473107, -0.02153038, -0.025279451, -0.010242999, 0.018156216, -0.025413347, 0.0036218707, 0.005111457, -0.014487483, -0.0059784297, -0.013690805, -0.171279, -0.0037222921, 0.01626829, -0.010417062, -0.0007322405, -0.001834367, 0.008776844, -0.012867348, -0.005884703, -0.0027147292, 0.022306973, 0.0042244, -0.049300287, -0.0157461, 0.016054058, 0.002781677, 0.00197161, 0.007980166, -0.014366977, -0.0071834885, 0.021048358, -0.024971493, 0.017955374, -0.007692291, 0.0043683373, 0.018557902, 0.01570593, 0.0027063608, 0.0011791164, -0.03698191, -0.014875779, 0.008455494, 0.016536081, 0.009486489, -0.001415107, 0.002960762, -0.008368462, -0.021878509, -0.022454258, 0.004686339, 0.012392019, 0.04394447, 0.016121006, -0.0068085813, 0.014085797, -0.0022946324, 0.008509053, -0.0063868104, 0.022333752, -0.026591627, 0.006497274, -0.01454104, 0.0080337245, -0.0059014396, 0.01602728, 0.02651129, -0.010738411, 0.014567819, -0.010303251, -0.031010175, -0.03821375, -0.0056403438, -0.00006835988, -0.0011732584, -0.021945456, -0.011146792, -0.023498643, 0.021409875, -0.026712133, -0.004190926, 0.002542339, 0.0062462203, -0.004522317, -0.02967122, 0.008334989, 0.00029415145, -0.018544514, 0.022240026, -0.024261847, -0.021811562, -0.020566333, 0.0390439, -0.025466906, 0.014059017, 0.013476572, -0.007451279, -0.0101760505, -0.021918677, 0.004093852, -0.003772503, 0.034304, -0.029483767, -0.025574021, -0.015893385, -0.003407638, 0.030233582, 0.007799407, 0.00002280406, 0.021838339, -0.01633524, -0.006875529, -0.010229609, 0.0053256894, 0.02011109, -0.010885696, 0.04016862, 0.028760733, 0.015183738, 0.013061496, -0.0073307734, 0.0077324593, 0.007739154, 0.015344413, 0.03783884, -0.012124228, 0.0145276515, -0.00086027797, 0.0006744981, 0.035375167, -0.0044620642, 0.030903058, 0.01567915, -0.0053189946, -0.014045628, -0.018852472, 0.0035683124, -0.09554776, -0.013791226, -0.015116791, 0.0013891648, -0.026377395, 0.019147042, -0.008254652, 0.040623866, -0.01656286, 0.01948178, 0.01310836, -0.006025293, 0.005971735, -0.0051348885, 0.019843299, 0.02007092, -0.027421778, 0.00007709446, 0.0038896615, 0.005737418, 0.010095714, -0.0044988855, 0.011294077, -0.001899641, -0.01567915, 0.007216962, -0.02095463, 0.024797428, -0.0064805374, 0.010691548, 0.01208406, -0.012867348, -0.0057775867, -0.023110347, -0.019588897, 0.0060821986, -0.019374665, -0.0061391043, 0.031331524, -0.018490955, 0.004043641, 0.0032017739, -0.003973346, -0.014982895, 0.008696507, -0.025989097, 0.007156709, 0.013523435, -0.0041139363, -0.03055493, -0.02793058, -0.011106623, -0.02851972, 0.023753043, 0.04689017, 0.0035850494, 0.009834617, 0.0096003, 0.016147785, 0.019856688, 0.0031582578, 0.004666255, -0.00829482, 0.0395527, -0.01077858, -0.020512775, -0.020512775, 0.012057281, 0.027006702, -0.021999015, -0.009633774, 0.02878751, -0.026645185, -0.005057899, -0.016964547, 0.003315585, -0.02910886, -0.008107367, 0.0138046155, -0.023538811, -0.0028804247, -0.03491992, 0.0076789013, -0.03781206, 0.014032238, 0.019642456, 0.021798171, -0.0074780583, -0.01602728, -0.011909996, -0.015183738, 0.0031063734, 0.0016686714, -0.036553446, -0.0018594724, 0.015906774, -0.009225393, 0.006755023, 0.0065776114, 0.0139117325, -0.0045524435, 0.0051583205, -0.049166393, 0.018477565, -0.010182745, 0.0031398472, 0.022614934, 0.0048905294, 0.027234325, -0.005191794, 0.026966535, -0.0012477378, -0.029483767, 0.010303251, -0.0072370465, 0.015505088, -0.015183738, -0.009948429, 0.00054185797, -0.016844042, 0.0015339392, -0.008495663, 0.01105976, 0.008375158, 0.013992069, 0.00698934, 0.0035448808, 0.01427325, 0.0080538085, 0.005382595, -0.021677665, 0.004900572, 0.008977687, -0.034812804, 0.005998514, 0.024984881, 0.0032687215, -0.02795736, 0.009124972, 0.0022778956, 0.0038126716, 0.012646421, 0.0019180516, -0.0128004, 0.013034717, -0.046033237, -0.00021506949, -0.005104762, -0.010309946, 0.0054093744, 0.01632185, -0.005737418, 0.016937768, 0.010945949, -0.018129438, 0.0039532618, -0.0047432445, -0.04051675, 0.03703547, 0.007551701, 0.0031264576, 0.00073935365, 0.012887432, 0.00020000625, 0.003869577, -0.012961075, -0.010443841, 0.038481537, 0.0037089025, -0.013643941, 0.03639277, -0.040329296, -0.022293584, 0.004087157, 0.011709153, 0.014902558, -0.006122367, 0.007852965, 0.003081268, 0.018571293, -0.0077190697, 0.020927852, 0.021195643, -0.00010554723, -0.029055303, 0.006269652, 0.029885454, 0.0060554193, -0.0075583956, 0.0008188541, 0.013041412, -0.006453758, -0.03467891, 0.005814408, 0.015090012, 0.010383588, 0.013818005, 0.018504344, -0.025761476, -0.011856438, 0.0052219206, 0.021423263, 0.00829482, -0.009881481, -0.01326234, -0.0038093242, -0.016951159, -0.005590133, -0.0067115068, -0.03483958, -0.010838833, 0.01717878, 0.038053073, -0.015612204, -0.004231095, 0.008027029, -0.008040419, 0.025667747, -0.005677165, -0.016455745, -0.028010918, 0.024342183, 0.010095714, 0.014219692, 0.016910989, -0.00083266204, 0.010182745, 0.021516992, 0.011950164, -0.03055493, -0.012104144, 0.004468759, -0.006969256, -0.014393755, -0.021342928, 0.0085626105, -0.015665762, 0.0021841687, -0.004234442, 0.030715605, -0.017138612, 0.042712633, 0.0062462203, 0.020057531, 0.008073892, -0.0326437, 0.01250583, 0.024342183, 0.04747931, -0.027020091, 0.0019414834, 0.015170349, -0.016362019, 0.02825193, -0.009566827, -0.039954387, -0.00697595, -0.023927107, -0.0285465, -0.000100316945, -0.024850987, 0.022963062, 0.002122242, 0.027877023, 0.0012870695, -0.018182995, -0.0079266075, 0.016174564, -0.010068934, -0.015090012, -0.0054227635, 0.0051516257, -0.013235561, -0.0075583956, -0.0131485285, 0.039365247, 0.0065575275, -0.011474836, 0.0028268667, -0.004425243, -0.0020703576, -0.010631295, -0.011702458, -0.0038394507, 0.0059784297, 0.032268792, 0.02244087, -0.023458473, -0.0053859423, -0.01925416]
    })
    .with_limit(2)
    .with_additional(["distance"])
    .do()
)

print(json.dumps(response, indent=2))

Limit the results​

You can set a limit on:

  • the number of results returned (with limit), or
  • how similar the results are to the query (with distance), or
  • the number of "jumps" in distance from the query (with the autocut filter).

autocut can be combined with limit, to set the maximum number of results returned by autocut.

Number of results​

You can set the maximum number of results returned with limit in the same way as shown in the search basics how-to guide.

Similarly, you can retrieve a maximum n objects after the first m results by using limit with offset as shown in the search basics how-to guide.

To limit the number of results returned by a near... query, add the limit operator. To start at a given offset, add the offset operator. For example if we want to obtain the animals in movies #2 and #3 from the nearText example above, we'll need to use offset: 1, limit: 2. The example below searches the JeopardyQuestion class for objects best matching "animals in movies", skips 1 object (offset) and returns the next 2 objects:

response = (
    client.query
    .get("JeopardyQuestion", ["question", "answer"])
    .with_near_text({
        "concepts": ["animals in movies"]
    }).with_additional(["distance"])
    .with_limit(2)
    .with_offset(1)
    .do()
)

print(json.dumps(response, indent=2))

Distance threshold​

You can set a threshold for similarity search by setting a maximum distance. The distance indicates how dissimilar two objects are.

Multiple distance metrics are available in Weaviate. You can set it in the schema as shown here.

The example below searches the JeopardyQuestion class for objects best matching "animals in movies", returning those with a distance less than 0.18:

max_distance = 0.18
response = (
    client.query
    .get("JeopardyQuestion", ["question", "answer"])
    .with_near_text({
        "concepts": ["animals in movies"],
        "distance": max_distance
    })
    .with_additional(["distance"])
    .do()
)

print(json.dumps(response, indent=2))
Why 0.18?

The numerical value for distance will depend on many factors, including the vectorization model and the distance metric used. As such, there are no hard and fast rules. In this case, we selected this value as our trial and error evaluation of this dataset indicated this value to produce relatively intuitive outputs.

Using certainty possible for cosine distance metric only

If the distance metric is set as cosine the certainty variable can be used, which normalizes the complement of distance to a value between 0 and 1.

Autocut​

Another way to limit the results returned by a similarity search is to use the autocut filter. Autocut takes a positive integer parameter N, looks at the distance between each result and the query, and stops returning results after the Nth "jump" in distance. For example, if the distances for six objects returned by nearText were [0.1899, 0.1901, 0.191, 0.21, 0.215, 0.23] then autocut: 1 would return the first three objects, autocut: 2 would return all but the last object, and autocut: 3 would return all objects.

Autocut can be used as follows:

response = (
    client.query
    .get('JeopardyQuestion', ['question', 'answer'])
    .with_near_text({
        'concepts': ['animals in movies'],
        'distance': max_distance
    })
    .with_autocut(1)
    .with_additional(['distance'])
    .do()
)

print(json.dumps(response, indent=2))
Example response

It should produce a response like the one below:

{
  "data": {
    "Get": {
      "JeopardyQuestion": [
        {
          "answer": "meerkats",
          "question": "Group of mammals seen <a href=\"http://www.j-archive.com/media/1998-06-01_J_28.jpg\" target=\"_blank\">here</a>:  [like Timon in <i>The Lion King</i>]",
          "_additional": { "distance": 0.17602634 }
        },
        {
          "answer": "dogs",
          "question": "Scooby-Doo, Goofy & Pluto are cartoon versions",
          "_additional": { "distance": 0.17842108 }
        }
      ]
    }
  }
}

Group results by a property or cross-reference​

Available from version v1.19

You can group search results by any arbitrary property or cross-reference.

The example below searches the JeopardyQuestion class for objects best matching "animals in movies", fetching the 10 closest results. Then those results are grouped by round, returning a maximum of two groups, each group with a maximum of two results (hits):

Grouping by cross-references

To group results by a cross-reference, try replacing the path value from round to hasCategory in the example below.

max_groups = 2
max_objects_per_group = 2
response = (
    client.query
    .get("JeopardyQuestion")
    .with_near_text({
        "concepts": ["animals in movies"]})
    .with_limit(10)
    .with_group_by(
        ["round"],
        groups=max_groups,
        objects_per_group=max_objects_per_group
    )
    .with_additional([
        """
        group {
            id
            groupedBy {
                path
                value
            }
            count
            minDistance
            maxDistance
            hits {
                question
                answer
            }
        }
        """
    ])
    .do()
)

print(json.dumps(response, indent=2))
Example response

It should produce a response like the one below:

{
  "data": {
    "Get": {
      "JeopardyQuestion": [
        {
          "_additional": {
            "group": {
              "count": 2,
              "groupedBy": {
                "path": [
                  "round"
                ],
                "value": "Jeopardy!"
              },
              "hits": [
                {
                  "answer": "meerkats",
                  "question": "Group of mammals seen <a href=\"http://www.j-archive.com/media/1998-06-01_J_28.jpg\" target=\"_blank\">here</a>:  [like Timon in <i>The Lion King</i>]"
                },
                {
                  "answer": "dogs",
                  "question": "Scooby-Doo, Goofy & Pluto are cartoon versions"
                }
              ],
              "id": 0,
              "maxDistance": 0.17842054,
              "minDistance": 0.17602539
            }
          }
        },
        {
          "_additional": {
            "group": {
              "count": 1,
              "groupedBy": {
                "path": [
                  "round"
                ],
                "value": "Double Jeopardy!"
              },
              "hits": [
                {
                  "answer": "fox",
                  "question": "In titles, animal associated with both Volpone and Reynard"
                }
              ],
              "id": 1,
              "maxDistance": 0.18770188,
              "minDistance": 0.18770188
            }
          }
        }
      ]
    }
  }
}

Add a conditional (where) filter​

You can add a conditional filter to your search results using the where argument.

The example below searches the JeopardyQuestion class for the top 2 objects best matching "animals in movies", as long as their round property is exactly "Double Jeopardy!":

response = (
    client.query
    .get("JeopardyQuestion", ["question", "answer", "round"])
    .with_near_text({
        "concepts": ["animals in movies"]
    })
    .with_limit(2)
    .with_additional(["distance"])
    .with_where({
        "path": ["round"],
        "operator": "Equal",
        "valueText": "Double Jeopardy!"
    })
    .do()
)

print(json.dumps(response, indent=2))
Example response

It should produce a response like the one below:

{
  "data": {
    "Get": {
      "JeopardyQuestion": [
        {
          "_additional": {
            "distance": 0.18759078
          },
          "answer": "fox",
          "question": "In titles, animal associated with both Volpone and Reynard",
          "round": "Double Jeopardy!"
        },
        {
          "_additional": {
            "distance": 0.19532347
          },
          "answer": "Swan",
          "question": "In a Tchaikovsky ballet, Prince Siegfried goes hunting for these animals & falls in love with 1 of them",
          "round": "Double Jeopardy!"
        }
      ]
    }
  }
}

Find least similar results​

Sometimes you may want to find objects that are the least similar to a given input. This might be possible for some distance metrics:

  • For cosine distances, perform a similarity search for a negative of a vector to find least.
  • For Euclidean or dot distances, the definition of "least similar" vector is not as clear-cut.

Accordingly, we generally recommend using cosine distance for this use case, and searching for a negative of your input vector with nearVector.

Further discussions

Here, the concept of least similar relates to finding vectors that are opposite to each other in the embedding space.

This may not necessarily mean that these least similar results have the opposite meaning in a semantic sense, such as antonyms in words.

Take the words rain and drought for example. While these are opposite concepts, both of them are unrelated to astrophysics. As such, in many models the distance between embeddings for 'rain' and 'astrophysics' will be likely greater than the distance between embeddings for 'rain' and 'drought'. Accordingly, you should consider the context of your use case when interpreting the results.

More Resources​

For additional information, try these sources.

  1. Frequently Asked Questions
  2. Weaviate Community Forum
  3. Knowledge base of old issues
  4. Stackoverflow
  5. Weaviate slack channel