AI Vector Search： Powering the Modern Enterprise [LRN1191].pdf

《AI Vector Search： Powering the Modern Enterprise [LRN1191].pdf》由會員分享，可在線閱讀，更多相關《AI Vector Search： Powering the Modern Enterprise [LRN1191].pdf（103頁珍藏版）》請在三個皮匠報告上搜索。

1、AI Vector Search:Powering the Modern EnterpriseOracle CloudWorld Copyright 2024,Oracle and/or its affiliates1Aurosish MishraSenior Director of Development,Data and AI,OracleRishin MitraPrincipal Oracle Cloud Technical Consultant,Version 1This presentation is intended to outline our general product d

2、irection.It is intended for information purposes only,and may not be incorporated into any contract.It is not a commitment to deliver any material,code,or functionality,and should not be relied upon in making purchasing decisions.The development,release,timing,and pricing of any features or function

3、ality describedfor Oracles products may change and remains at the sole discretion of Oracle Corporation.Statements in this presentation relating to Oracles future plans,expectations,beliefs,intentions,and prospects are“forward-looking statements”and are subject to material risks and uncertainties.A

4、detailed discussion of these factors and other risks that affect our business is contained in Oracles Securities and Exchange Commission(SEC)filings,including our most recent reportson Form 10-K and Form 10-Q under the heading“Risk Factors.”These filings are available on the SECs website or on Oracl

5、es website at http:/ information in this presentation is current as of September 2024 and Oracle undertakes no duty to update any statement in light of new information or future events.Some regulatory certifications or registrations to products or services referenced herein are held by Cerner Corpor

6、ation.CernerCorporation is a wholly owned subsidiary of Oracle.Cerner Corporation is an ONC-certified health IT developer and a registered medical device manufacturer in the United States and other jurisdictions worldwide.The materials in this presentation pertain to Oracle Health,Oracle,Oracle Cern

7、er,and Cerner Enviza which are all wholly owned subsidiaries of Oracle Corporation.Nothing in this presentation should be taken as indicating that any decisions regarding the integration of any EMEA Cerner and/or Enviza entities have been made where an integration has not already occurred.Oracle Clo

8、udWorld Copyright 2024,Oracle and/or its affiliates2Forward-Looking statementsThe following is intended to outline our general product direction.It is intended for information purposes only,and may not be incorporated into any contract.It is not a commitment to deliver any material,code,or functiona

9、lity,and should not be relied upon in making purchasing decisions.The development,release,timing,and pricing of any features or functionality described for Oracles products may change and remains at the sole discretion of Oracle Corporation.The materials in this presentation pertain to Oracle Health

10、,Oracle,Oracle Cerner,and Cerner Enviza which are all wholly owned subsidiaries of Oracle Corporation.Nothing in this presentation should be taken as indicating that any decisions regarding the integration of any EMEA Cerner and/or Enviza entities have been made where an integration has not already

11、occurred.Oracle CloudWorld Copyright 2024,Oracle and/or its affiliates3Safe harbor statementCopyright 2024,Oracle and/or its affiliates4Databases have traditionally performed value-based searchesFind revenue by productsCopyright 2024,Oracle and/or its affiliates5There is a growing volume of unstruct

12、ured business data that must be searched by semantics or meaningFind products that match a photo or descriptionCopyright 2024,Oracle and/or its affiliates6Vector Search is a breakthrough capability for searching data by semanticsCopyright 2024,Oracle and/or its affiliates7What is Vector Search?Vecto

13、r SearchCopyright 2024,Oracle and/or its affiliates8A new technology that enhances information retrieval by mapping queries to relevant data in your database based on semantics,instead of precise matches,using vectors to measure similaritySimilarity SearchVectors are used in AI to capture the semant

14、ics of data:Images,documents,videos,or even structured dataA vector is a sequence of numbers,called dimensions,used to capture the important“features”of the dataRepresent the semantic content of data,not the actual words in a document or pixels in an imageProduced by AI/ML Embedding ModelsVector3342

15、162150Copyright 2024,Oracle and/or its affiliates9Laptop Gen 32Support Rep:Jane DoeEmail:Severity 1Spontaneous rebootResolvedApplied OS Update 42Support IncidentAn Example Business Scenario:The vector for a support incident could be.Each dimension(number),represents a different feature of the suppor

16、t incidentVectorFeaturesSeverityCurrent OS versionSymptomsProductStatus3342162150Copyright 2024,Oracle and/or its affiliates10Note:Features are often chosen by ML algorithms and are not as simple as shown hereThe main operation on vectors is the Mathematical Distance between them82133262Distance(Euc

17、lidean Squared)=(3-2)2+(1-6)2+(2-2)2+(8-3)2)There are many mathematical distance formulas(e.g.,Euclidean,Cosine,Hamming)Copyright 2024,Oracle and/or its affiliates11Support incident vectors when collapsed into 2 dimensions instead of hundreds could look like thisd1Support Incidents for Laptops runni

18、ng slowlySupport Incident for Desktop crashingd2Copyright 2024,Oracle and/or its affiliates12Similarity Property:Support Incidents that are more similar also produce vectors that are closer togetherd1Support Incidents for Laptops running slowlySupport Incident for Desktop crashingd2Not as SimilarSim

19、ilarCopyright 2024,Oracle and/or its affiliates13The more similar two entities are,the shorter the distance between their vectorsThe Similarity Property powers AI Vector SearchCopyright 2024,Oracle and/or its affiliates14User QueryTop K matchesAI Vector SearchML Embedding ModelQuery VectorcompareDat

20、a Corpus(encoded with the same embedding model)Arent there dozens of Vector Databases out there already?Copyright 2024,Oracle and/or its affiliates15Why is Oracle adding AI Vector Search?One solution is to continuously send business data to a vector databaseBusiness DatabaseEnterprises get most valu

21、e by combining semantic searches with value-based business data searchesCopyright 2024,Oracle and/or its affiliates16Vector Database42162150Support Incident DataOne solution is to continuously send business data to a vector databaseBusiness DatabaseCopyright 2024,Oracle and/or its affiliates17Vector

22、 Database42162150You need to send a lot of data since you cant predict the question that might be askedEnterprises get most value by combining semantic searches with value-based business data searchesSupport Incident Data Product Data Customer Data One solution is to continuously send business data

23、to a vector databaseBusiness DatabaseCopyright 2024,Oracle and/or its affiliates18Vector Database42162150Enterprises get most value by combining semantic searches with value-based business data searchesSupport Incident Data Product Data Customer Data Causes data staleness,adds complexity,compromises

24、 security One solution is to continuously send business data to a vector databaseBusiness DatabaseCopyright 2024,Oracle and/or its affiliates19Vector Database42162150Enterprises get most value by combining semantic searches with value-based business data searchesSupport Incident Data Product Data Cu

25、stomer Data Enterprise DBs typically have an order of magnitude more sophisticatedquery capabilities,fault-tolerance,security,etc.,than niche Vector DBsConverged AI DatabaseThe best solution is to add AI Vector Search to your business databaseCombine semantic searches with value-based searches to ex

26、tract sophisticated insightsThere is no need to move and synchronize data,manage multiple products,etc.Copyright 2024,Oracle and/or its affiliates20Better Together:Business Data and Business Vectors Where to use AI Vector Search?Find Similar Support TicketsBiometric pattern recognitionFind Similar P

27、roducts Product RecommendationDetect manufacturing anomaliesEnterprise Similarity Search Use-CasesCopyright 2024,Oracle and/or its affiliates21Natural language catalog searchCATALOGCopyright 2024,Oracle and/or its affiliates22AI Vector Search and Generative AIWith the Power of Generative AI,Vector S

28、earch allows you to interact with business data using Natural Language Copyright 2024,Oracle and/or its affiliates23Using LLMs to answer my questions|Easy-Peasy!A text response is sent back to the user in whatever format was specified123Becauseit is bad.Maybe buya new oneWhy does my laptop keep rebo

29、oting?UserCopyright 2024,Oracle and/or its affiliates24GenAI uses the users question plus its general knowledge to generate an answerUsers natural language questions is sent to LLMLLMThe LLMs response based on its general knowledge isnt great.Why?Copyright 2024,Oracle and/or its affiliates25Becausei

30、t is bad.Maybe buya new oneWhy does my laptop keep rebooting?LLMUserUsing LLMs to answer my questions|Hmmm.WHAAAT!Large Language Models LLMs are trained on a broad range of data from the internetHowever,they are frozen on a past snapshot of the internet with no access to private enterprise dataLLMs

31、can also hallucinate(i.e.,inaccurate responses)LLMs by themselves therefore often provide poor-quality responses to support questionsCopyright 2024,Oracle and/or its affiliates26Becauseit is bad.Maybe buya new oneWhy does my laptop keep rebooting?LLMUserUsing LLMs to answer my questions|Not Enough b

32、y ItselfAI Vector Search improves Generative AI by augmenting prompts with private database content This helps produce better answers to user questionsAvoid having to train LLMs on sensitive enterprise data Called:Retrieval Augmented Generation(RAG)Copyright 2024,Oracle and/or its affiliates27A user

33、s natural language question is encoded as a vector and sent to AI Vector Search1RAG with AI Vector Search works like thisCopyright 2024,Oracle and/or its affiliates28Why does my laptop keep rebooting?334216215AI Vector SearchAI Vector Search finds private content such as incidents that match the use

34、rs questionCan search for relevant product type and customer location2Incident ReportsProduct and customer dataUserThe users question is augmented with relevant content3LLMLLM uses the question plus the content and general knowledge to provide a better informedanswer4The issue is withthe firmware co

35、ntrolling the fan.Apply OS update 42 while plugged in,and in a cool air-conditioned environment to prevent overheatingUserRAG with AI Vector Search is very powerfulCopyright 2024,Oracle and/or its affiliates29AI Vector SearchThe LLMs response is much-improved thanks to AI Vector Search and the addit

36、ional contextThe issue is withthe firmware controlling the fan.Apply OS update 42 while plugged in,and in a cool air-conditioned environment to prevent overheatingIncident ReportsProduct and customer dataLLMCopyright 2024,Oracle and/or its affiliates30AI Vector Search:Key FeaturesAI Vector Search Hi

37、ghlightsGenerate vector embeddings from unstructured data Store vectors in table columns using new VECTORtypeBuild approximate vector indexes on VECTORcolumnsPerform AI Vector Search on VECTORcolumns using SQLIntegrate with Mission-Critical Enterprise CapabilitiesCopyright 2024,Oracle and/or its aff

38、iliates31GenerateStoreIndexSearchIntegrateCopyright 2024,Oracle and/or its affiliates32GenerateVectorsAI Vector Search offers 3 alternatives for vector embedding generationUse Pre-Created EmbeddingsLoad vectors directly from external files into database into VECTOR columns or map the data as externa

39、l tablesUse an external embedding serviceGenerate embeddings using external callouts via UTL_TO_EMBEDDING()PLSQL function in the DBMS_VECTOR packageUse a database resident embedding modelGenerate embeddings using the VECTOR_EMBEDDING()SQL function using an imported ONNXembedding model so that no dat

40、a leaves the databaseOracle CloudWorld Copyright 2024,Oracle and/or its affiliates33Vector Embedding Generation|Your Way123Copyright 2024,Oracle and/or its affiliates34StoreVectors35CREATE TABLE Support_Incidents(id NUMBER,incident_textCLOB,incident_vecVECTOR(768,FLOAT32);Optional#of dimensionsOptio

41、naldimension formatCopyright 2024,Oracle and/or its affiliatesVECTOR Datatype to Store and Process VectorsNEW VECTOR DatatypeINSERT INTO Support_Incidents(1,Problem.,TO_VECTOR(1.1,2.2,.)SELECT FROM_VECTOR(incident_vector)FROM Support_Incidents;Insert VectorsFLOAT32,FLOAT64,INT8,BINARY Fetch Vectors1

42、.1,2.2,3.3Native VECTOR support available for Python,Node.js,JDBC,OCI,ODP.Net driversCopyright 2024,Oracle and/or its affiliates36IndexVectorsCopyright 2024,Oracle and/or its affiliates37NEW vector indexes provide 100 x faster and highly accurate similarity searches Graph-based index where vertices

43、represent vectors and edges between vertices represent similarityHierarchical Navigable Small Worlds(HNSW)index gold standard of Vector IndexingIn-Memory only index-highly efficient for both accuracy and speedVector Index|Neighbor Graph Vector IndexCopyright 2024,Oracle and/or its affiliates38Layer

44、3Layer 2Layer 1Layer 0Partition Vector Index(e.g.,IVF_FLAT index)Partition-based index with vectors clustered into table partitions based on similarityEfficient scale-out index for unlimited data sizeVector Index|Neighbor Partition Vector IndexCopyright 2024,Oracle and/or its affiliates39Vector Inde

45、x CreationCopyright 2024,Oracle and/or its affiliates40Basic index creation syntax:CREATE VECTOR INDEX incident_idx ON SUPPORT_INCIDENTS(incident_vector)ORGANIZATION INMEMORY NEIGHBOR GRAPH|NEIGHBOR PARTITIONSDISTANCE COSINE|EUCLIDEAN|MANHATTAN|.TARGET_ACCURACY|Choosing the ORGANIZATIONfor an index

46、is simple:If the index data will fit in-memory,use INMEMORY NEIGHBOR GRAPH,else use NEIGHBOR PARTITIONSThe DISTANCEclause is optional(the default is COSINE)TARGET_ACCURACY:Specify the default accuracy(recall)of queries when using the indexSimple for users to specify accuracy than low-level index cre

47、ation parametersCopyright 2024,Oracle and/or its affiliates41SearchVectorsVector Search SQL|Distance FunctionCopyright 2024,Oracle and/or its affiliates42NEW SQL Function to compute distance between vectors to gauge similarityDifferent embedding models can use different distance metrics,but the basi

48、c concept remains the same:The Distance between two vectors is smaller for entities that are more similarDistance functions supported in 23ai are:COSINE(Default),EUCLIDEAN,EUCLIDEAN_SQUARED,HAMMING,MANHATTAN,DOTVECTOR_DISTANCE(VECTOR1,VECTOR2,)Vector Search SQL|Specifying Similarity SearchCopyright

49、2024,Oracle and/or its affiliates43Find the top 10 matching support incidentsSupport Incident Search ExampleSELECT FROMSupport_IncidentsORDER BY VECTOR_DISTANCE(incident_vector,:search_vector)FETCH FIRST 10 ROWS ONLY;Vector Search SQL|Specifying Target AccuracyCopyright 2024,Oracle and/or its affili

50、ates44Find the top 10 matching support incidentsNEWTARGET ACCURACY clause to specify desired accuracy of approximate similarity searchOverrides accuracy specified at index creationSELECT FROMSupport_IncidentsORDER BY VECTOR_DISTANCE(incident_vector,:search_vector)FETCH FIRST 10 ROWS ONLY TARGET ACCU

51、RACY|Vector Search SQL|Combine value-based and semantic searchCopyright 2024,Oracle and/or its affiliates45Find the top 10 matching support incidents that were filed within the last 7 daysSELECT FROMSupport_IncidentsWHEREIncident_Date SYSDATE 7ORDER BY VECTOR_DISTANCE(incident_vector,:search_vector)

52、FETCH FIRST 10 ROWS ONLY;Value-based Attribute Filters can be combined seamlessly with Vector Search in SQLOptimizer picks the best access plan based on filter selectivityHIGH=PRE-FILTERMEDIUM=IN-FILTERLOW=POST-FILTER*Coming Soon!Vector Search SQL|Similarity Search over JoinsCopyright 2024,Oracle an

53、d/or its affiliates46Find the top 10 matching support incidents fora Laptop reported by customers in Las VegasSELECT FROMSupport_Incidents SIJOINProducts P ON SI.product_id=P.idJOINCustomers C ON SI.customer_id=C.idWHEREP.Type=LaptopAND C.City=Las VegasORDER BY VECTOR_DISTANCE(SI.incident_vector,:se

54、arch_vector)FETCH FIRST 10 ROWS ONLY;Combines customer and product data,and AI search in a few lines of SQLEssential capability as enterprise data is normalizedAny developer or DBA can learn to use it in 10 minutesCopyright 2024,Oracle and/or its affiliates47IntegrateAI Vector Search is fully integr

55、ated with Oracle Database 23ai featuresCopyright 2024,Oracle and/or its affiliates48Parallel ExecutionAnalyticsShardingPartitioningSecurityRACTransactionsExadataAI Vector Search powers Complete Gen AI pipelineData SourcesDocument LoadersDocument Transformation(e.g.,Text Splitting,Summarization)Embed

56、ding ModelsVector DatabaseSimilarity SearchLLMsUserAI Vector Search in Oracle 23ai DatabaseRAG1.Native Database APIs for Gen-AI,from Document Processing to REST call-outs to LLM Providers2.Tight integration with 3rdparty frameworks such as LangchainCopyright 2024,Oracle and/or its affiliates49Copyri

57、ght 2024,Oracle and/or its affiliates50Real-World Use-CaseMeet RossOracle CloudWorld Copyright 2024,Oracle and/or its affiliates5253Oracle CloudWorld Copyright 2024,Oracle and/or its affiliatesProblem Statement|AI Vector Semantic SearchOracle CloudWorld Copyright 2024,Oracle and/or its affiliates54V

58、ector SearchIn23aiCopyright 2024,Oracle and/or its affiliates55Application ArchitectureOracle CloudWorld Copyright 2024,Oracle and/or its affiliates56Architecture|AI Vector Semantic SearchUserBrowserORDS21.47149APEXOracle Database 23ai with Oracle APEX24.1Copyright 2024,Oracle and/or its affiliates5

59、7Oracle APEX ApplicationOracle CloudWorldCopyright 2024,Oracle and/or its affiliates58APEX Application|AI Vector SearchCreate Region Add Text Field Oracle CloudWorldCopyright 2024,Oracle and/or its affiliates59APEX Application|AI Vector SearchAdd ButtonOracle CloudWorldCopyright 2024,Oracle and/or i

60、ts affiliates60APEX Application|AI Vector SearchCreate Interactive ReportOracle CloudWorldCopyright 2024,Oracle and/or its affiliates61APEX Application|AI Vector SearchType=SQL QueryOracle CloudWorldCopyright 2024,Oracle and/or its affiliates62APEX Application|AI Vector SearchAI Vector Search Query

61、goes in hereOracle CloudWorldCopyright 2024,Oracle and/or its affiliates63APEX Application|AI Vector SearchOracle CloudWorldCopyright 2024,Oracle and/or its affiliates64APEX Application|AI Vector SearchCopyright 2024,Oracle and/or its affiliates65DatabaseConfigurationStepsOracle CloudWorld Copyright

62、 2024,Oracle and/or its affiliates66Step 1Step 2Step 3Step 4Step 5Step 6Create UserProvide necessary GrantsLoad Augmented Model(all-MiniLM-L12-v2 Model in ONNX Format)to Oracle Database 23aiCreate Table/Insert Data into Table and Generate Vector EmbeddingsGenerate Vector Embedding for Search string(

63、Optional)Create Vector IndexDatabase Setup Steps|AI Vector SearchDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates67User CreationCREATE USER rishoradev IDENTIFIED BY Welcome123QUOTA UNLIMITED ON USERS;User created.SQLDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/o

64、r its affiliates68Grant User Access to Import Mining ModelGRANT DB_DEVELOPER_ROLE,CREATE MINING MODEL TO rishoradev;Grant succeeded.SQLDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates69Download the ONNX model Database Setup|AI Vector SearchCopyright 2024,Oracle and/or its a

65、ffiliates70Load Augmented ModelBEGINDBMS_VECTOR.LOAD_ONNX_MODEL(directory =DM_DUMP,file_name=all_MiniLM_L12_v2.onnx,model_name=ALL_MINILM_L12_V2);END;/PL/SQL procedure successfully completed.SQLDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates71Validate the ModelSELECTmodel_

66、name,algorithm,mining_functionFROMuser_mining_modelsWHEREmodel_name=ALL_MINILM_L12_V2;SQLMODEL_NAME ALGORITHM MINING_FUNCTION_ _ _ALL_MINILM_L12_V2 ONNX EMBEDDINGDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates72Simple TestSELECT VECTOR_EMBEDDING(ALL_MINILM_L12_V2 USING I L

67、ove Las Vegas as DATA)AS embedding;SQL-2.90911552E-002,-8.46567824E-002,2.63756663E-002,3.70936394E-002,4.80906107E-003,8.88874196E-003,7.3078312E-002,6.34469092E-003,1.13900416E-002,3.47795784E-002,-1.76432729E-002,-1.12383179E-002,-4.84913066E-002,2.34519858E-002,3.46485227E-002,-4.02011424E-002,4

68、.67650704E-002,2.89806677E-002,.VECTOR_EMBEDDING(ALL_MINILM_L12_V2 USING I Love Las Vegas as DATA)I Love Las Vegas7149Input StringInput StringVector Embedding ModelVector Embedding ModelVectorVectorOracle CloudWorld Copyright 2024,Oracle and/or its affiliates73Database Setup|AI Vector SearchCopyrigh

69、t 2024,Oracle and/or its affiliates74Data LoadDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates75Create TableDESCXXRM_REPORT_LISTName Null?Type_ _ _REP_ID NOT NULL NUMBERREP_TYPE VARCHAR2(20)PRODUCT_FAMILY VARCHAR2(4000)REP_CUST VARCHAR2(50)REP_TECH VARCHAR2(20)REPORT_NAME V

70、ARCHAR2(4000)REP_DESCRIPTION VARCHAR2(4000)REP_DESC_VEC VECTOR(*,*)REPORT_NAME_VEC VECTOR(*,*)SQLDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates76Create TableDESCXXRM_REPORT_LISTName Null?Type_ _ _REP_ID NOT NULL NUMBERREP_TYPE VARCHAR2(20)PRODUCT_FAMILY VARCHAR2(4000)REP_

71、CUST VARCHAR2(50)REP_TECH VARCHAR2(20)REPORT_NAME VARCHAR2(4000)REP_DESCRIPTION VARCHAR2(4000)REP_DESC_VEC VECTOR(*,*)REPORT_NAME_VEC VECTOR(*,*)SQLVECTOR Data TypeDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates77Generate VECTOR EmbeddingSQLUPDATExxrm_report_list xsrlSETxs

72、rl.report_name_vec=(SELECT TO_VECTOR(VECTOR EMBEDDING(ALL_MINILM_L12_V2 usingreport_name as data)FROMXXRM_REPORT_LIST xrWHERExr.Rep_Id=xsrl.Rep_Id);Copyright 2024,Oracle and/or its affiliates78AI Vector Search in ActionDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates79Perfo

73、rming Semantic SearchSELECT xsrl.product_family,xsrl.report_nameFROM xxrm_report_list xsrlORDER BY VECTOR_DISTANCE(TO_VECTOR(VECTOR_EMBEDDING(ALL_MINILM_L12_V2 USING subledger accounting as data),rep_desc_vec)FETCH FIRST 5 ROWS ONLY;SQLOracle CloudWorldCopyright 2024,Oracle and/or its affiliates80PR

74、ODUCT_FAMILY REPORT_NAME_ _Accounting Hub Subledger Accounting Method Setups ReportAccounting Hub Journal Ledger ReportAccounting Hub Create Multiperiod Accounting Execution ReportAccounting Hub Activate Subledger Journal Entry Rule Set Assignments Execution ReportGeneral Ledger Account Analysis for

75、 Contra Account ReportElapsed:00:00:00.46Database Setup|AI Vector SearchPipeline|AI Vector Semantic SearchOracle CloudWorldCopyright 2024,Oracle and/or its affiliates8123aiVector EmbeddingVector EmbeddingObject EmbeddingObject EmbeddingSearch for Similar Object VectorsSearch for Similar Object Vecto

76、rsClient InputClient Input(Images,Documents,Texts,Sound,etc.)(Images,Documents,Texts,Sound,etc.)Search MatchesSearch Matches-|Id|Operation|Name|Rows|Bytes|TempSpc|Cost(%CPU)|Time|-|0|SELECT STATEMENT|5|71400|1021 (1)|00:00:01|*1|COUNT STOPKEY|2|VIEW|1207|16M|1021 (1)|00:00:01|*3|SORT ORDER BY STOPKE

77、Y|1207|3973K|4840K|1021 (1)|00:00:01|4|TABLE ACCESS FULL|XXRM_REPORT_LIST|1207|3973K|170 (0)|00:00:01|-Oracle CloudWorldCopyright 2024,Oracle and/or its affiliates82Execution PlanDatabase Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates83INDEX CreationCREATE VECTOR INDEX XXRM_REP_DE

78、SC ONXXRM_REPORT_LIST(REP_DESC_VEC)ORGANIZATION INMEMORY NEIGHBOR GRAPHDISTANCE COSINE WITH TARGET ACCURACY 95;Vector INDEX created.SQLEntry PointLayer 3Layer 2Layer 1Layer 0Query VectorOracle CloudWorldCopyright 2024,Oracle and/or its affiliates84PRODUCT_FAMILY REPORT_NAME_ _Accounting Hub Subledge

79、r Accounting Method Setups ReportAccounting Hub Journal Ledger ReportAccounting Hub Create Multiperiod Accounting Execution ReportAccounting Hub Activate Subledger Journal Entry Rule Set Assignments Execution ReportGeneral Ledger Account Analysis for Contra Account ReportElapsed:00:00:00.34Oracle Cl

80、oudWorldCopyright 2024,Oracle and/or its affiliates85-|Id|Operation|Name|Rows|Bytes|Cost(%CPU)|Time|-|0|SELECT STATEMENT|5|71400|2 (50)|00:00:01|*1|COUNT STOPKEY|2|VIEW|5|71400|2 (50)|00:00:01|*3|SORT ORDER BY STOPKEY|5|16855|2 (50)|00:00:01|4|TABLE ACCESS BY INDEX ROWID|XXRM_REPORT_LIST|5|16855|1 (

81、0)|00:00:01|5|VECTOR INDEX HNSW SCAN|XXRM_REP_DESC|5|16855|1 (0)|00:00:01|-Improved PerformanceCopyright 2024,Oracle and/or its affiliates86SQL BEGINSELECTVECTOR_EMBEDDING(ALL_MINILM_L12_V2 USING subledger accounting AS data)INTO:query_vec From Dual;END;/SQL SELECTxsrl.product_family,xsrl.report_nam

82、eFROMxxrm_report_list xsrlORDER BY VECTOR_DISTANCE(:query_vec,rep_desc_vec)FETCH FIRST 5 ROWS ONLY;Elapsed:00:00:00.00Improved PerformancePL/SQL procedure successfully completed.SQL variable query_vec vector;Copyright 2024,Oracle and/or its affiliates87SQL SELECT elapsed_time,executions,sql_textFROM

83、V$SQL WHERESQL_TEXT like SELECT%xxrm%;SQL SELECT xsrl.product_family,xsrl.report_nameFROMxxrm_report_list xsrlORDER BY VECTOR_DISTANCE(:query_vec,rep_desc_vec)FETCH FIRST 5 ROWS ONLY;SQL SELECT elapsed_time,executions,sql_textFROMV$SQL WHERESQL_TEXT like SELECT%xxrm%;ELAPSED_TIME EXECUTIONS SQL_TEXT

84、-11824 6 SELECT xsrl.product_family,xsrl.report_nameFROM xxrm_report_listxsrl ORDER BY VECTOR_DISTANCE(:query_vec,rep_desc_vec)FETCH FIRST 5 ROWS ONLYELAPSED_TIME EXECUTIONS SQL_TEXT-12851 7 SELECT xsrl.product_family,xsrl.report_nameFROM xxrm_report_listxsrl ORDER BY VECTOR_DISTANCE(:query_vec,rep_

85、desc_vec)FETCH FIRST 5 ROWS ONLY12851-11824=1027 1ms Database Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates88Error encountered while creating INDEXError starting at line:1 in command-CREATE VECTOR INDEX XXRM_REP_DESC ON XXRM_REPORT_LIST(REP_DESC_VEC)ORGANIZATION INMEMORY NEIGHBOR

86、 GRAPHDISTANCE COSINEWITH TARGET ACCURACY 95Error report-ORA-51961:The vector memory area is out of space.Database Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates89Problemshow parameter vector_memory_size;SQLNAME TYPE VALUE-vector_memory_sizebig integer 0Database Setup|AI Vector Se

87、archCopyright 2024,Oracle and/or its affiliates90Solution(Version 23.5.0.24.07)ALTER system SET vector_memory_size=200m SCOPE=spfile;SQLSQLNAME TYPE VALUE-vector_memory_sizebig integer 208Mshow parameter vector_memory_size;Database Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates91C

88、ombining Value-Based Search with Semantic SearchSELECT xsrl.product_family,xsrl.report_nameFROMxxrm_report_list xsrlWHEREPRODUCT_FAMILY=General LedgerORDER BY VECTOR_DISTANCE(TO_VECTOR(VECTOR_EMBEDDING(ALL_MINILM_L12_V2 USING subledger accounting as data),rep_desc_vec)FETCH FIRST 5 ROWS ONLY;SQLOrac

89、le CloudWorldCopyright 2024,Oracle and/or its affiliates92PRODUCT_FAMILY REPORT_NAME_ _General Ledger Account Analysis for Contra Account ReportGeneral Ledger Journals ReportGeneral Ledger General Ledger and Subledger Accounting by Journal Lines ReportGeneral Ledger Journals Batch Summary ReportGene

90、ral Ledger Journals Day Book ReportElapsed:00:00:00.550Database Setup|AI Vector SearchCopyright 2024,Oracle and/or its affiliates93AI Vector Search:Roadmap94Key InnovationsExadata InnovationsDistance Computation Offload,BINARY+INT8HNSW IndexTransactions,RAC,Persistence,Memory Sizing AdvisorLLMs and

91、FrameworksOllama,LlamaIndex,LangChainJS,Langchain4JHybrid IndexesGeneral ImprovementsPartition-Local Indexes,Auto Accuracy Calibration,Jaccard Distance,Reranking APIsMiscellaneousPost-Filtering,Sparse VectorsCopyright 2024,Oracle and/or its affiliates.6COMING SOONOracle ML IntegrationVECTOR output f

92、rom Feature Extraction algorithmsDatabase FreeOptimized IVF Index Space UtilizationCopyright 2024,Oracle and/or its affiliates95AI Vector Search:Natural Language to SQLCopyright 2024,Oracle and/or its affiliates96Use Natural Language totalk to your databaseSelect AISpeaks HumanConverse with your Dat

93、a|Select AICopyright 2024,Oracle and/or its affiliates97Use a standard SELECT statement followed by AI and your questionProcess the result as you would any other SQL result setKeep refining your questions until you get the most desirable answerCopyright 2024,Oracle and/or its affiliates98Key Takeawa

94、ysCopyright 2024,Oracle and/or its affiliates99Oracle AI Vector Search powers the Modern AI EnterpriseSeamlessly integrated with mission-critical capabilities of Oracle Database 23aiPerform sophisticated information retrieval by combining semantic searches with value-based searches on your business

95、dataConverse with business data in Natural Language using RAGBuild modern AI apps for the enterprise natively in the databaseAI Vector Search|How to get startedCopyright 2024,Oracle and/or its affiliates100https:/ CloudWorld Copyright 2024,Oracle and/or its affiliatesTry Everythingfor FREEAI Solutio

96、ns HubOracle LiveLabsOracle Database F these steps to earn your free swag!1.Scan the QR code and complete our survey2.Visit us at the Demogrounds3.Show your survey confirmation4.Claim your FREE SWAG!We value your feedback!Oracle CloudWorld Copyright 2024,Oracle and/or its affiliates102of people love free swag!Aurosish Mishra()Rishin Mitra()Thank YouOracle CloudWorld Copyright 2024,Oracle and/or its affiliates103