SCORE Semantic Search
Advanced Retrieval-Augmented Generation (RAG) powered by the SCORE Genkit microservice.
SCORE Semantic Search uses the Google Genkit framework to process CMS content into vector embeddings. The service chunks unstructured text, generates embeddings via a configured model, and synchronizes the results with a vector database to enable Retrieval-Augmented Generation (RAG) workflows.
Text Extraction & Embedding
The microservice extracts and segments text from Drupal nodes, taxonomy, and PDF attachments. These segments are processed through an embedding model to compute high-dimensional vectors representing their semantic structure before being indexed into the database.
Vector Database Synchronization
The generated high-dimensional embeddings are synchronized into dedicated vector databases like Qdrant. This foundational layer allows the system to perform high-speed nearest-neighbor similarity searches, enabling the AI to retrieve context based on mathematical proximity.
GraphRAG Knowledge Traversal
By integrating vectors directly into Neo4j's property graph, the system unlocks advanced GraphRAG capabilities. This allows the AI agent to not only find semantically similar text but also traverse structured relationships between entities, enabling highly deterministic and interconnected context retrieval.
OpenTelemetry Tracing
The Genkit framework is instrumented with OpenTelemetry to monitor the complete RAG pipeline. It records individual trace spans for text extraction, embedding generation, and vector retrieval, exposing exact execution latencies and API token usage to the observability stack.
What framework does SCORE Semantic Search use for processing CMS content into vector embeddings?
SCORE Semantic Search utilizes the Google Genkit framework to process CMS content into vector embeddings.
How does SCORE Semantic Search prepare unstructured text for embedding generation?
What types of content does the microservice extract text from for embedding?
Which vector database is mentioned for synchronizing generated embeddings?
What capability does integrating vectors into Neo4j's property graph unlock?
How does the system enable AI agents to traverse structured relationships between entities?
What observability is provided by instrumenting the Genkit framework with OpenTelemetry?
What specific metrics are exposed to the observability stack due to OpenTelemetry tracing?
What is the primary purpose of synchronizing embeddings into a vector database?
What is the benefit of using GraphRAG capabilities with Neo4j?
How does the system ensure deterministic and interconnected context retrieval?
What is the role of the vector database in the SCORE Semantic Search workflow?