Skip to main content

PhariaAI v1.251100.0 – Release Notes

Release Versionv1.251100.0
Release DateNovember 5, 2025
AvailabilityOn-premise and hosted
important

Known issue: Reference display may appear incorrectly when working with documents in PhariaAssistant - the team is investigating and preparing a fix.

PhariaAssistant​

This release brings major enhancements to PhariaAssistant’s knowledge management capabilities, introducing Public and Private Collections to expand data control and reusability as well as strengthening cross-team collaboration

Public and Private Collections: Empower users with persistent knowledge bases​

AvailabilityOn-premise and hosted

This release introduces Public and Private Collections, allowing users to create, manage, and reuse document collections directly within Assistant. The new capability enables users to store and organize knowledge sources for long-term use, removing the need for repetitive uploads and empowering teams to share organizational knowledge at scale.

  • Knowledge workers can create and manage Private Collections by uploading or syncing files, explore their contents, and use both public and private collections seamlessly in Assistant Chat;
  • Admins can create, update, and delete both Public and Private Collections, manage files synced from SharePoint, and control visibility across the organization.

Collections can be created by uploading local files (.pdf, .docx, .txt) or by connecting directly to SharePoint for continuous synchronization.

Key Benefits:

  • Persistent, reusable context: Users can store frequently used documents and access them anytime in Assistant, improving workflow efficiency;
  • Admin-managed knowledge sharing: Admins can set up Public Collections accessible across the organization - no API or technical setup required;
  • Seamless SharePoint integration: Upload or sync content from Microsoft SharePoint with automatic authentication handling and metadata visibility.
note

This release does not include advanced access control (e.g., group sharing), document versioning, or connectors beyond SharePoint and direct file uploads.

PhariaStudio​

This release introduces Search Stores in Data Platform - a new, scalable search architecture that enhances performance, simplifies configuration, and enables enterprise-grade retrieval and indexing across multimodal datasets.

AvailabilityOn-premise and hosted

In this release, we are introducing a new abstraction for managing and retrieving searchable content named Search Store API while continuing to support the previous, Collections API. The new Search Store API aims to improve performance and scalability, while enhancing the developer experience. Unlike the previous system of namespaces and collections, Search Stores simplify setup and bring search infrastructure closer to production-ready deployments. Each Search Store is self-contained, with its own index, making it easier to manage, scale, and optimize for enterprise workloads. A Search Store continues to offer the same embedding, indexing and metadata filtering capabilities as Collections API.

Key Capabilities:

  • Simplified architecture: Each store now maintains a single index, reducing configuration complexity and aligning with production search patterns;

  • Enhanced scalability: Supports high-performance retrieval across large, multimodal datasets;

  • Support for pagination at all levels;

  • Expanded indexing capabilities: Ingest and index multiple documents with customizable, searchable fields;

  • Indexes: Controlling Search Behavior

    • Define how documents are chunked, embedded, and queried;
    • Continue supporting the configuration: chunk size, overlap, and embedding types (symmetrical or asymmetrical);
    • Optimize for recall, precision, and query performance.

  • Advanced Embedding Options:

    • Introducing multiple embedding strategies in PhariaSearch for flexible retrieval. You can now tailor embeddings to data structure, query patterns, and performance needs;
    • Support for OpenAI-compatible embeddings that extends the flexibility by supporting industry-standard embedding models. This is supported only by Search stores API and not in Collections API.

  • Improved Search Features:

    • Metadata-based filtering (e.g. labels, timestamps);
    • Relevance-based ranking for more accurate results.
note

We will continue to maintain the support for Namespaces, Collections, and Indexes in the Collections API to ensure backwards compatibility within the Data platform until 31st December 2025. Together with the new Document model and Search Store architecture, we are providing a unified, and scalable foundation for building enterprise-grade semantic search experiences in the Data platform.

Other updates​

  • PhariaStudio has received improvements to the evaluation feature, introducing filtering, searching, and sorting capabilities for enhanced usability and analysis.

For PhariaOS:

PhariaOS enabled support for customizing metrics export paths and ports in PhariaOS deployments. Users can now define which endpoint their application metrics are exposed on, allowing seamless integration with custom observability setups and Prometheus scrapers.

If not specified, PhariaOS defaults to the /metrics path and a 30-second scraping interval. These parameters can now be overridden in the deployment request to better fit application needs:

"scrapingInterval": "5s"      # default: 30s  
"path": "/custom-metrics"     # default: /metrics

For PhariaInference API:

The team has introduced new metrics to track translation endpoint activity across all language pairs in PhariaAI.

Early insights suggest that translation usage is sparse across the full language matrix, with most activity focused on key pairs such as X <-> German and X <-> English. These metrics will help identify high-frequency pairs, prioritize resource allocation, and support scalability as user adoption grows.