Gregius Data
The orchestration layer for AI workflows in WordPress
AI interactions are rapidly becoming the default means of navigating digital interfaces, where digital assets serve as powerful beacons signaling sources of truth. Because AI is still emerging, its implementation promotes observability, tuning, and transparency.
Orchestration Control Plane
A centralized backend control plane designed to manage, monitor, and configure your entire AI infrastructure directly within the WordPress dashboard. This unified registry decouples complex pipeline engineering from your theme logic, allowing you to manage provider connections, index database syncs, track real-time telemetry logs, and maintain absolute custody of your data environment from a single interface.
Agentic RAG Assistant
A production-ready Gutenberg block that delivers a secure, grounded, and context-aware conversational interface directly to your site visitors. Fully integrated into the native WordPress editor, the block functions as a highly configurable orchestration node, allowing you to select dedicated connection layers, route traffic through agentic and reranking models, enforce system prompts, and enable real-time response streaming within a zero-SaaS, self-hosted framework.
Orchestration
WordPress remains the source of truth, while PostgreSQL runs alongside as a secondary execution layer for vector operations. This architecture provides your AI with semantic memory and embeddings to enable fast, context-aware retrieval, allowing your data and operations to stay safely inside your own infrastructure.
Models
Multiplatform intelligence demands seamless flexibility across the modern AI ecosystem.
The unified registry centralizes access to Language models, Embedding models, and Reranking models. This infrastructure enables you to orchestrate diverse providers fluidly, keeping your system adaptable and fully decoupled from single vendor limitations.
Connections
Modern data architectures require fluid agility to scale and migrate workloads without friction.
A central PostgreSQL registry delivers native support for PDO and REST providers to easily facilitate multiple databases and remote portability. This configuration ensures your execution layer remains fully decoupled, allowing you to route data dynamically while maintaining structural flexibility across any environment.
Sync
Effective semantic retrieval relies on precise boundary control over the data fed into your vector pipeline.
Granular content selection settings allow you to define exactly what gets synchronized per connection, filtering by specific post types, and publishing statuses. This localized control gives you total command over your indexing footprint, ensuring your AI only consumes relevant data while optimizing your storage infrastructure.
Vectors
Maximizing the accuracy of semantic search requires a flexible approach to generating high fidelity embeddings.
Deep integrations offer support for embedding APIs, such as OpenAI, Gemini, and DeepSeek. There are also algorithmic alternatives like TF-IDF and Hashing-TF, which do not have token costs. This hybrid capability allows control over budget and pipeline design. It enables the balancing of cloud models with zero cost local execution.
Prompts
A resilient AI application requires total control over the context and instructions passed to your models.
A central prompt management system decouples your logic from hardcoded templates, facilitating real-time hydration, versioning, and conditional formatting. This structure allows you to build sophisticated, context aware prompt pipelines that adapt dynamically to user interactions while ensuring consistent execution.
Logs
Maintaining a resilient pipeline requires deep visibility into how your system executes instructions in real time.
An observable stream captures every AI workflow interaction, tracing the telemetry between users, editors, and models. This transparent ledger delivers complete diagnostic control, allowing you to audit system behavior, refine prompt accuracy, and troubleshoot bottlenecks directly inside your environment.
Features
Crafted at the intersection of two distinct architectural paradigms, these features embrace both WordPress and AI patterns to bridge proven infrastructure stability with emerging technological innovation.
Abilities
Register engine capabilities natively with the WordPress Abilities API and seamlessly expose them as machine-discoverable tool contracts for language models.
Interactions
Capture search and Retrieval Augmented Generation (RAG) events as persistent records, delivering the foundational stream needed for operational analysis, debugging, and downstream data workflows.
Providers
Support multiple database backends and cloud AI services through unified, extensible interfaces, preventing single-vendor lock-in while maintaining complete architectural flexibility.
RAG (Retrieval Augmented Generation)
Provide grounded question-answering over native site content using highly configurable retrieval, generation, and governance pipelines.
RAG Evals (evaluations)
Generate framework-ready evaluation datasets from pre-collected interaction prompts, enabling systematic, evidence-based measurement of your LLM response quality.
RAG Benchmarks
Execute repeatable, evidence-backed validation of RAG pipeline behavior to ensure release readiness and proactive regression detection.
Retry-queue
Automate recovery from transient database sync failures using smart error classification, exponential backoff scheduling, and background dead-letter management.
Search
Deliver fully integrated retrieval across lexical, typo-tolerant, and semantic matching protocols while preserving native operational safety and fallback behavior.
Tools
Expose a stable, multi-provider tool selection mechanism designed specifically to power complex, multi-step retrieval-augmented workflows.
WP CLI
Expose a comprehensive set of operator command-line tools to automate sync, vector generation, answering, listing, and log management workflows.
Hooks
A standardized system of actions and filters provides a stable, discoverable, and governable extension surface for developers.
Questions, answered
Want to know more? Check out these answers to common questions.
Gregius Data is an open-source orchestration layer and control plane designed to manage, monitor, and configure AI workflows directly within WordPress. It decouples complex AI pipeline engineering from your theme logic, allowing site owners to run advanced AI interactions—like semantic search and custom assistants—while keeping their operational data securely self-hosted.
Gregius Data is built for enterprise developers, web agencies, and serious WordPress operators who want to integrate AI without relying on expensive SaaS platforms or risking data privacy. It is ideal for organizations requiring absolute custody over their data environments, custom prompt pipelines, and multi-provider flexibility.
No. Gregius Data functions within a zero-SaaS, self-hosted framework. Your data and operations stay safely inside your own infrastructure using WordPress as the source of truth, alongside a secondary PostgreSQL execution layer for vector operations.
Yes. The architecture separates complex processing from standard site operations by offloading vector storage and embeddings to a dedicated PostgreSQL database. It features built-in retry queues with exponential backoff and a robust WP-CLI toolset to safely automate high-volume data syncs and vector generation in the background.
Users can control the vector indexing footprint. The central control plane allows defining which site content is synchronized to the vector database by filtering data by post types, categories, or publishing statuses.
The platform supports premium language models, embeddings, and reranking APIs, such as OpenAI, Gemini, and DeepSeek. It also includes algorithmic, zero-cost local execution options like TF-IDF and Hashing-TF.
Yes. Gregius Data uses native WordPress developer paradigms. It includes action and filter hooks, a WP-CLI interface for automation, and an Abilities API that registers custom site engine capabilities as machine-discoverable tool contracts for LLMs.
Gregius Data features a standardized system of developer action and filter hooks designed for deep architectural customization. Developers can leverage these hooks to build entirely new plugins that extend the core orchestration layer. For example, the premium add-on, Gregius Intelligence, was built using these exact hooks to introduce an advanced continuous improvement layer—adding tools for interaction history, journey continuity, and editorial curation dashboards.
Yes. The plugin includes RAG Evaluation and RAG Benchmarking modules. These generate datasets from interaction logs, allowing repeatable validations to test LLM response quality before updates.
Next Steps
View on GitHub: You can review, fork, and inspect the entire codebase and core logic over at the repository on GitHub.
Gregius Data is the open-source AI orchestration layer for WordPress.
Sign up for the newsletter
Practical insights on how real-world AI adoption is reshaping the open web.
