Building Next-gen Search With OpenSearch
summary
OpenSearch 3.x exposes agentic RAG: register a model (connector + model group), create a plan-execute-reflect agent with conversation index memory, and bind tools like the cat indices tool, index mapping and search (vector DB + ML model) or external services via MCP external server. Demo connects an OpenAI connector, registers a model, queries recipe and product indices for a faukatcha cost estimate, successfully retrieves items and prices but mis-scales quantities and misses some retail-format pricing, showing capabilities and current limitations.
AI-written summary · about
chapters (8)
- 0:00Agentic RAG and OpenSearch
- 3:39Limitations of standard RAG
- 7:57From prompting to agentic RAG
- 12:31Standardizing with MCP
- 14:44OpenSearch AI journey
- 21:57Demo: e-commerce agent
- 30:35Demo: MCP server connection
- 34:13Conclusion
related talks
26mOpenSearch: Conversations with Context
25mKeynote: OpenSearch in the Agentic Era
40mAgentic Relevance Tuning With OpenSearch: Faster, Easier, and Higher Quality Search at Scale
34mImplementing an open-source RAG with OpenSearch
28mAI Agents in Action: Harnessing MCP and OpenSearch for Modern Use Cases
06mSponsored Keynote: OpenSearch - See Everything: Open Observability for Agentic AI
26mRevolutionizing Search: Unveiling Conversational Search
42mBuilding Self-Healing Infrastructure: Agentic AI & Vector Search in OpenSearch