I have relativelly big dataset 300K records
in form
825292676,Whole Grain Bread - Var51,“Whole grain sandwich bread, 24 oz loaf”,Grocery,Hy-Vee,1234567890-051,Hy-Vee,BREAD-WG-24-051,HV-BREAD-24-051,GKDGREH83LQVZ57G
2745019024,Brake Rotor - Rear - Var18,“OEM replacement brake rotor for rear wheels, vented”,Automotive,AutoZone,AZ-BRK-REAR-018,Duralast,DL-REAR-RT-018,DL-BRK-REAR-018,6I6LTMWRF1TDS4BS
6609301978,Breadboard - 830 Tie Points - Var19,830-point solderless breadboard for prototyping,Electronics,Sparkfun,PRT-00112-019,Sparkfun,PRT-00112-019,SF-PRT-00112-019,CBKIVEJD1LV2RWNI
1702666664,Motor Oil 5W-30 Synthetic - Var5,“Full synthetic motor oil 5W-30, 5 qt.”,Automotive,Walmart,WM-5W30-5QT-005,SuperTech,ST-5W30-005,WM-ST5W30-005,CW3AYCUWZNPB3H2D
2918778670,Amazon Basics Alkaline Batteries AA - 20 Pack - Var17,Long-lasting alkaline AA batteries for household and office use,Electronics,Amazon,AB-AA20PK-017,Amazon Basics,AMZ-AA20-017,AMZ-AA20PK-017,NJMYTYTEHWQQQ1JP
3757049369,Vibrador recargable con control remoto - Var21,“Silicona médica, sumergible, 10 modos de vibración”,Adultos,Lovehoney.es,LH-VIB-REMOTE-021,Lovehoney,LH-RMT-VIB-021,LH-VIB-RMT-021,A6YJHB0QS2PGIM7E
4417732063,Crown Royal Whisky - Var5,750ml bottle of Canadian whisky,Beverages,Diageo,DIAG-CR750-005,Crown Royal,CR-750ML-005,CR750-DIAG-005,0ABPPO9JLUA002FM
8456972672,Arduino Uno R3 - Var29,Arduino Uno R3 development board with ATmega328P microcontroller,Electronics,Arduino,A000066-029,Arduino,A000066-029,ARD-UNO-R3-029,JWMSSD0R7OYPA91S
9205784349,Arduino Uno R3 - Var39,Arduino Uno R3 development board with ATmega328P microcontroller,Electronics,Arduino,A000066-039,Arduino,A000066-039,ARD-UNO-R3-039,WVUDAQRK3GG7SL2F
8789150140,Organic Milk - 1 Gallon - Var27,1 gallon of whole organic milk,Grocery,Target,TG-MILK-1G-027,Good & Gather,MILK-ORG-1-027,TG-GG-MILK1-027,JV2BXS0K1IS0H0PU
8210253584,Caja de herramientas metálica - Var121,“Caja organizadora con cierre metálico, 5 compartimentos”,Herramientas,BricoDepot,BD-TOOLBOX-MET-121,Dexter,DX-BOX5-121,BD-CJTOOL-121,6SC8XRR77RQPWR4R
2983582667,Raspberry Pi 4 Modelo B - 4GB - Var37,“Placa base con procesador ARM Cortex-A72, 4 GB RAM, puertos USB y micro-HDMI”,Electrónica,PcComponentes,RPI4-MOD4G-PC-037,Raspberry Pi Foundation,RPI4-4GB-037,PC-RPI4-4GB-037,EL1WJ6MKEF3L7FA6
5750090061,Raspberry Pi 4 Modelo B - 4GB - Var68,“Placa base con procesador ARM Cortex-A72, 4 GB RAM, puertos USB y micro-HDMI”,Electrónica,PcComponentes,RPI4-MOD4G-PC-068,Raspberry Pi Foundation,RPI4-4GB-068,PC-RPI4-4GB-068,1X5G9TVE7BELR35O
2008410498,Raspberry Pi 4 Model B - Var5,“Raspberry Pi 4 Model B, 4GB RAM”,Electronics,Mouser,358-PI4B-4GB-005,Raspberry Pi,PI4B-4GB-005,RPI-4B-4GB-005,XP9ZP2E5C85PI5N8
6265202783,Interruptor doble de pared - Var105,“Interruptor de luz blanco, montaje en pared, 10A”,Electricidad,Leroy Merlin,INT-DOBLE-LM-105,Lexman,LEX-INT-DB-105,LM-INT-2-105,O0UTB9MQGGODYIEC
5989197328,Breadboard - 830 Tie Points - Var8,830-point solderless breadboard for prototyping,Electronics,Sparkfun,PRT-00112-008,Sparkfun,PRT-00112-008,SF-PRT-00112-008,DBX4XLSROT4HVUVB
6373998355,Industrial Nitrile Gloves - Large - Var46,“Powder-free, 5-mil thick nitrile gloves, box of 100”,Industrial,Uline,S-14264-046,Uline,S-14264-046,UL-S-14264-046,Y8RXDG0RHIAE1PF6
210791325,Organic Milk - 1 Gallon - Var3,1 gallon of whole organic milk,Grocery,Target,TG-MILK-1G-003,Good & Gather,MILK-ORG-1-003,TG-GG-MILK1-003,2S2ALNYYJIHLBTF0
8631936180,Raspberry Pi 4 Model B - Var41,“Raspberry Pi 4 Model B, 4GB RAM”,Electronics,Mouser,358-PI4B-4GB-041,Raspberry Pi,PI4B-4GB-041,RPI-4B-4GB-041,HX3RV4T6C8I5A4GA
8378802215,1μF Ceramic Capacitor - 50V X7R - Var25,Surface-mount MLCC 1μF 50V ±10% X7R dielectric,Electronics,DigiKey,445-1UF-X7R-50V-025,Murata,GRM188R71H105KA93D-025,GRM1UF50VX7R-025,AZGG09U6DFJEF37D
2561923138,Motor Oil 5W-30 Synthetic - Var26,“Full synthetic motor oil 5W-30, 5 qt.”,Automotive,Walmart,WM-5W30-5QT-026,SuperTech,ST-5W30-026,WM-ST5W30-026,2A5EEF2ZTJNX3DE5
different kind of product, electronics, food, beverages etc.
with query
”milk” and k = 5 it returns number of ”Organic Milk - …..” records
with k < 5 it returns few records “Cafe moulu arabica -….“
{
"table": "products_openai",
“_source”: [“title”],
“size”: 5,
“knn”: {
“field”: “vector”,
“query”: “arduino”,
“k”: 500
}
}
If I try to found “Arduino” ( i have such records in db )
it returns “Transistore bipolare NPN silicio” with k<500!
and only from k=500 it returns “Arduino Uno R3 - Var7”
Is it any guidance how to choose this k param?