1、Monitoring Monitoring PgVectorPgVector RAG Agentic RAG Agentic ApplicationsApplicationsJayita Bhattacharyya,Jayita Bhattacharyya,Data ScientistData ScientistHOW2025 PostgreSQL&IvorySQL Eco ConferenceRAG&RAG&friendsfriendsCONTENTSCONTENTSHOW2025 PostgreSQL&IvorySQL Eco ConferencePostgres to Postgres
2、to pgvectorpgvectorAgents&Agents&ObservabilityObservabilityBonus-Bonus-PgaiPgai pgvectorscalepgvectorscaleDemoDemoNext Steps ResourcesNext Steps ResourcesTraditional DB vs Vector DB Traditional DB vs Vector DB RAG WorkflowRAG WorkflowPgevctorPgevctor-pgvectorisaPostgreSQLextensionthataddsthevectorda
3、tatypeandfunctionsforvectoroperationstoaPostgreSQLdatabase.Its key features include:Vector Data TypeVector Similarity Search,Distance Metrics,Indexing Language AgnosticHOW2025 PostgreSQL&IvorySQL Eco Conferencepgvector-pythonisaPythonlibraryforpgvector.ItallowsyoutousepgvectorwithSQLAlchemy,makingit
4、easytointegratepgvectorintoyourexistingPythonapplicationsWhatareAIAgentsHas AGI been achieved?Has AGI been achieved?Yes&NoIve been having hard time vibe“debugging”!LocaltoProduction-MLModelsScalability:Deployed models should handle high volumes of requests efficiently.Scalability ensures smooth oper
5、ation even under heavy loads.Performance:Models must provide accurate and timely predictions.Throughput&Latency:Optimize for low latency and high throughput.MLSystemMonitoringMonitoring involves systematically collecting and analyzing data to assess the performance of a system,process,or device over
6、 time.Purpose:It helps track predefined metrics and alerts when thresholds are crossed.Monitoring is proactive and aims to prevent issues.Data Focus:Monitoring tools work with predetermined data sets,narrowing the analytical frame.Example:Setting up alerts for CPU usage exceeding a certain threshold