Secure Vector Search
Our connectors pick up documents from your content sources and make sure that the vector search index stays in sync. Prior to indexing a configurable document processing pipeline is used to extract textual contents from binary documents, to tag these or to split the texts into sections, which are indexed separately.
This data thus becomes indexed in your vector search. Our connectors also provide access control information, as well as user-group relationships. In turn, you get a early binding security trimmed vector search.
Query Pipelines Making Your Integrations Easy
Our query pipelines combine secure search with generative AI. You can use them via our APIs, for instance from search applications, bots or within custom plugins or leverage our search interface.
- After submitting a query, configured query transformers take care of intent recognitions, input language detections or keyword extractions. The resulting query, including an acl filter for allow and deny ACLs is sent to the vector search index.
- Afterwards our result transformer jump into place. Besides mapping fields, result transformers comprise LLM prompts which rephrase and combine the results.
- As a result, your application get the data which makes a difference to the daily work of your users.
Features
- Security-trimmed RAG
- Language detection
- Intent recognition
- Query transformation pipelines
- Tailored search queries
- Result transformation pipelines
- Gen AI integration
- APIs to integrate in your applications
- Tailored
- Easy to setup and configure
- Use RheinInsights and other connectors
- Integrations into the RheinInsights Search Interface
