Organizations are rapidly building AI-powered applications, from Retrieval-Augmented Generation (RAG) solutions and semantic search to intelligent assistants and recommendation engines. These applications all have one thing in common: they depend on fresh, continuously synchronized data and the vector embeddings that make it searchable and meaningful for AI models.
Keeping those embeddings up to date, however, often means building and maintaining an entirely separate pipeline after data replication. Operational data is replicated to its destination, then another service generates embeddings, introducing additional infrastructure, orchestration, and operational overhead.
Imagine a customer support application where support tickets are continuously replicated into a vector-enabled database to power semantic search or an AI assistant. Traditionally, generating embeddings would require a separate workflow after replication. With the latest Oracle GoldenGate 26ai capabilities, embedding generation can now become part of the replication pipeline itself, helping keep operational data and its AI-ready representation synchronized in real time.
That’s exactly the direction we’re taking with the latest OCI GoldenGate release.
This release adds AI Model connections and support for the new AI Service in Oracle GoldenGate 26ai, so vector embeddings can be generated during real-time replication and written to vector-enabled datastores such as Oracle AI Database 26ai, Oracle Autonomous AI Database, PostgreSQL, Snowflake, or Elasticsearch. We are also expanding OCI GoldenGate connectivity with new connection types and broader support for modern lakehouse architectures.
Bringing AI closer to where data moves
OCI GoldenGate has long helped organizations move mission-critical data reliably and in real time across heterogeneous databases, clouds, and streaming platforms. Whether supporting operational reporting, real-time analytics, cloud migrations, or event-driven architectures, GoldenGate has helped organizations keep data continuously synchronized across increasingly complex environments.
As AI becomes another major consumer of enterprise data, data integration platforms have an opportunity to do more than simply move data, they can help prepare it for AI workloads as it flows.
With this release, OCI GoldenGate is taking that next step.
The new AI Model connection type allows administrators to configure a secure connection to supported AI providers and associate it with a specific embedding model. OCI GoldenGate currently supports AI Model connections for:
- Oracle Cloud Infrastructure Generative AI
- OpenAI
- Google Gemini
- Voyage AI
Once an AI Model connection is configured and assigned, it enables Oracle GoldenGate 26ai’s new AI Service to generate vector embeddings during replication.
Rather than replicating data first and enriching it later, GoldenGate can now generate vector embeddings inline as data flows through the replication pipeline. This simplifies AI-ready architectures by reducing the need for separate embedding-generation workflows while helping ensure that operational data and vector representations remain synchronized.
The result is a cleaner architecture with fewer moving parts for organizations building AI applications on top of continuously changing enterprise data.
Building the foundation for AI-powered data integration
The initial release centers on vector embedding generation, a foundational building block for many modern AI applications. Embedding generation is the first AI capability integrated directly into OCI GoldenGate’s real-time replication engine. It establishes an architectural foundation for bringing AI services closer to where enterprise data moves.
As organizations continue adopting AI-native architectures, data integration platforms will increasingly be expected not only to replicate data between systems, but also to enrich, transform, and prepare that data for AI consumption as it flows.
AI Model connections and the AI Service provide a flexible foundation that can support increasingly intelligent real-time data integration scenarios as both Oracle GoldenGate and enterprise AI ecosystems continue to evolve.
Expanding connectivity across modern data platforms
While AI is the headline capability in this release, modern data architectures also depend on broad connectivity across enterprise databases, streaming platforms, and open lakehouse technologies. This update continues expanding OCI GoldenGate’s ecosystem to help customers integrate data wherever it resides.
Google Cloud Managed Service for Apache Kafka
OCI GoldenGate now supports Google Cloud Managed Service for Apache Kafka, making it easier to incorporate managed Kafka deployments running in Google Cloud into real-time replication and streaming architectures.
This addition further strengthens OCI GoldenGate’s multicloud capabilities by simplifying connectivity across cloud providers without requiring custom integration approaches.
Oracle Exadata Exascale
We’re also introducing a new connection type for Oracle Exadata Exascale.
As customers adopt Oracle’s latest database infrastructure, OCI GoldenGate continues to provide trusted real-time replication capabilities, allowing Exadata Exascale environments to integrate seamlessly with existing enterprise data architectures.
Enhanced Apache Iceberg support
OCI GoldenGate continues investing in open lakehouse architectures with enhancements to the Apache Iceberg connection type.
This release adds support for:
- OCI Object Storage
- Amazon S3 Tables
These additions provide greater flexibility for organizations building cloud-native data lakes and lakehouses while delivering real-time CDC into Apache Iceberg tables across multiple storage platforms.
Continuing the evolution of real-time data integration
Enterprise data architectures continue to evolve. Organizations are expanding across hybrid and multicloud environments, embracing open data formats, and increasingly building AI-powered applications that require continuously synchronized operational data.
The expanded connectivity options continue broadening the range of enterprise technologies that OCI GoldenGate can integrate in real time. At the same time, AI Model connections and the Oracle GoldenGate 26ai AI Service introduce the first AI-native capability within OCI GoldenGate’s replication engine, allowing vector embeddings to be generated as data is replicated. By bringing AI closer to where enterprise data moves, OCI GoldenGate is laying the foundation for a new generation of intelligent real-time data integration.
If you’re already using OCI GoldenGate to replicate operational data into analytics platforms, lakehouses, or AI-enabled databases, now is a great time to explore how embedding generation can become part of your existing replication pipelines. This release is an important step toward making AI a native part of real-time data integration, and we look forward to continuing that journey in future OCI GoldenGate releases.
To learn more about the Oracle GoldenGate 26ai AI Service and supported AI providers, see the Oracle GoldenGate 26ai announcement, read about AI with OCI GoldenGate, and go through the quickstart: Use AI Models in OCI GoldenGate to Create Vector Embeddings. For complete information about all newly supported connection types, refer to the OCI GoldenGate documentation.