This blog was co-authored by Jack Christie and Dave Stone.
The hype around generative AI (GenAI) continues—and yet Gartner predicts that 30% of GenAI initiatives will be abandoned after proof of concept by the end of 2025. Businesses are hitting roadblocks due to the complexity of piecing together disparate off-the-shelf AI components or they’re getting cold feet after seeing disappointing results from generic chatbots and copilots. But the real issue is more foundational: 60% of data leaders cite data governance as their top priority, which is at odds with the immaturity and public cloud-based nature of many AI tools.
This aligns with a general theme that’s emerging across industries: businesses are looking to streamline infrastructure and take back control of their data—with 70% of them looking to break down data silos and rein in data sprawl to power better business intelligence and more sustainable growth.
EDB is committed to meeting these enterprise needs with the EDB Postgres® AI platform. We’re enabling businesses to leverage existing infrastructure and skills to unlock advanced GenAI capabilities while retaining complete control of their data, and we’re providing faster time to insights with rapid analytics across data unified in trusted, familiar Postgres. On the heels of our Q4 2024 releases, let’s dive deeper into exactly how our latest AI and Analytics offerings solve these modern business challenges.
Now generally available: AI Accelerator
We’re excited to announce general availability of EDB AI Accelerator—the fastest way to test and launch enterprise GenAI inference applications with the powerful Pipelines extension that comes preloaded with pgvector. This enables robust management and automation of GenAI data orchestration, retrieval, and search.
Architecture: GenAI workloads on Postgres
The Pipelines extension comes with three key capabilities that make GenAI development remarkably simple:
- Managed Pipelines integrate embedding generation, storage, indexing, and management internally within the Pipelines extension. This includes an integrated model runtime that can be used out of the box, allowing for easy access to other models—just specify the model and then Pipelines takes care of the entire workflow from data ingestion to similarity search. Managed Pipelines coordinate the complete lifecycle of AI data processing, providing a streamlined, automated solution for transforming raw data into AI-ready resources within your Postgres environment.
- Auto Embedding automates the process of generating and updating embeddings for AI data stored in Postgres tables, ensuring your vector stores stay current without manual intervention. By continuously monitoring your database, Auto Embedding eliminates the need for manual embedding refresh, maintaining an always-up-to-date, AI-searchable index that reflects the latest state of your data.
- The Intelligent Retriever encapsulates all processing needed to make AI data searchable and retrievable through similarity. It fetches text and image data from Postgres tables or object storage with a simple AI retriever function call. The Intelligent Retriever abstracts away the complexities of vector similarity calculations, providing a streamlined interface for AI data within your familiar Postgres environment—transforming your Postgres database into a powerful AI search engine.
These capabilities of the Pipelines extension combine with the native vector compatibility provided by the included pgvector extension, resulting in a complete vector toolkit for management of embeddings in Postgres—all with just 5 lines of SQL code instead of the 130+ lines it would normally take. In addition, swappable configurations make it easy to switch between models, integrate multiple data modalities, and select from a variety of storage locations—in the cloud or on prem—enabling tailored performance and efficient data management that adapts to any need across your business. No more tearing down and rebuilding with each iteration. Transform your Postgres database into a powerful GenAI semantic search engine across that’s 4.22x faster than purpose-built vector databases.
Use cases powered by AI Accelerator
AI Accelerator delivers real-world GenAI value for our customers. Here’s a spotlight on a few key use cases.
Sovereign AI
Sovereign AI enables private GenAI where your data resides, ensuring complete control over AI operations. Run preferred models alongside existing data in Postgres, eliminating data movement and enhancing security. Remove cloud data access fees with flexible private deployment options and locally hosted models. Learn more about Sovereign AI.
Cognitive AI
EDB Postgres AI simplifies development of Cognitive AI applications that bring together multiple data modalities. It leverages the Pipelines extension to effortlessly generate, manage, and retrieve text and image embeddings for powerful similarity and semantic search. Learn more about Cognitive AI.
Virtual Expert
With the Pipelines extension, developers can use existing SQL skills to build enterprise-hardened Virtual Expert chatbots and copilots with efficient vector data management in Postgres. Robust compliance and security give peace of mind when training systems on proprietary data in highly regulated industries. Learn more about Virtual Expert.
See AI Accelerator in action
To learn how to start building with AI Accelerator, visit the getting started docs for the Pipelines extension.
Seeking an end-to-end AI solution? EDB and Griptape are partnering to provide ground-up enterprise GenAI solutions tailored to match your business needs. Learn about the partnership on our alliance page.
Now in tech preview: Analytics Accelerator
Scaling analytics workloads is crucial for modern enterprises that deal with high volumes of data and demand for timely insights, and they’re looking for a unified solution that minimizes extra operational overhead. Now available in tech preview, EDB Analytics Accelerator leverages lakehouse ecosystem integration and the PGAA extension to provide rapid analytics and intelligent data management in Postgres.
Architecture: Rapid analytics workloads on Postgres
The PGAA extension unlocks revolutionary capabilities for analytical workloads in Postgres:
- The Vectorized Query Engine is optimized for columnar data formats and provisioned on a separate, ephemeral, scale-to-zero compute resource, allowing analytical queries to run 30x faster, on average, compared to the standard row-based Postgres engine. Use standard SQL to write Postgres tables to 5x smaller columnar lakehouse table formats in external object storage—and query them just like any Postgres table.
- When PGAA is combined with AutoPartition in EDB Postgres Distributed (PGD), it provides Tiered Tables functionality. Query hot data and cold data with a dedicated node, ensuring optimal performance by automatically offloading cold data to columnar tables in object storage, reducing the complexity of managing analytics over multiple data tiers. Store lakehouse tables in flexible storage locations that are up to 18x more cost-efficient versus solid state drives—including Amazon S3, MinIO, or the local filesystem. You can choose between managed storage locations, with lifecycle and security fully managed for you, or bring your own.
Use cases powered by Analytics Accelerator
Analytics Accelerator supercharges highly extensible Postgres, enabling analytics use cases centered on unified business data.
Customer 360
Achieve comprehensive customer insights with transactional, analytical, and AI data stores unified in a single Postgres environment instead of fragmented across disparate systems—and query all of it with 30x the speed of standard Postgres. Directly impact the bottom line with faster, better decisions driven by holistic customer data. Learn more about Customer 360.
Dynamic Analytics
Query and view real-time and historical data on the fly without bloating storage costs or compromising performance. Tiered tables automatically offload less frequently used data into efficient columnar formats in object storage—drastically simplifying the complexity of managing multiple data tiers, boosting query speed, and improving cost efficiency. Learn more about Dynamic Analytics.
Operational Analytics
Run high-performance OLAP queries over operational data without impacting ongoing transactional workloads. Rather than competing for resources with OLTP processes, provision separate analytics nodes that scale up seamlessly to meet complex query demands and scale back down to zero when not in use — ensuring live business operations are uninterrupted. Learn more about Operational Analytics.
See Analytics Accelerator in action
To get started with Analytics Accelerator, request tech preview access.
Wrapping up
These releases are enabling our customers to confidently tackle GenAI initiatives, streamline infrastructure, and pursue more unified data strategies—all with enterprise control, security, and deployment flexibility of EDB Postgres AI. Stay tuned for continued innovation.
Interested in more details about Q4 releases across EDB? Read the Q4 2024 Release Highlights.