PhariaAI v1.250700.0 – Release Notes
| Release Version | v1.250700.0 |
| Release Date | July 2, 2025 |
| Availability | On-premise and hosted |
PhariaAssistant
In this release, we’re introducing foundational improvements across the Chat experience, evaluation capabilities, and interface navigation. Together, these updates aim to boost user efficiency, improve model assessment workflows, and enhance discoverability at scale — supporting enterprise growth and AI integration at every level.
Beta Chat now supports existing features in Chat such as documents and collections. Also, we will make the agentic technology powering the Beta Chat the default in future and intend to deprecate the existing Chat architecture latest by 03 September 2025.
Enhanced File Upload: Bigger, faster, more flexible uploads for document-heavy workflows
Upload more files, upload bigger files — and power richer conversations in Chat. With support for up to 10 files per session, 100MB per file, and up to 500 pages per document, this improvement removes critical limitations that previously blocked enterprise use cases.
Key Benefits
- Upload more files: Up to 10 files per conversation (previously 3);
- Upload larger files: Up to 100MB per file, with support for documents up to 500 pages (previously 20MB and 200 pages);
- Support enterprise use cases: Reduce workflow friction and enable deeper, document-rich interactions in Chat.
Upload allocation is still per conversation — no shared file repository or persistent uploads. Progress indicators are not yet token-based.
Unified Chat Interface: A modern, accessible, and streamlined Chat experience
The Chat interface has been fully redesigned — combining the best of previous implementations into a single, modern UI. This update simplifies usage, improves accessibility, and ensures all key features are seamlessly available in one place.
Key Benefits
- One seamless experience: Unified design combines file uploads, collection selection, and conversation tools in a single, modern interface;
- Fully accessible: Meets up-to-date accessibility standards;
- Simpler workflows: Centralized context controls make key actions easier to find and use.
This is Phase 1 of the UI unification. Some advanced agentic features, full theme customization, and advanced conversation management are still in development.
Agentic Chat Evaluation Framework: Enterprise-grade multi-turn LLM assessment
A powerful framework for evaluating large language models through multi-turn conversations — enabling consistent, automated, and extensible assessments of conversational AI systems. The framework is designed for enterprise-scale validation, with CLI/Python SDK access and full PhariaStudio integration, and introduces a configurable evaluation architecture with support for facilitator/critic roles, document-based Q&A assessments, and local/remote dataset handling.
Key Benefits
- Robust multi-turn evaluation: Assess LLMs against ground-truth facts using configurable agentic dialogues;
- Integrated developer tools: Run and manage evaluations via CLI or Python SDK for seamless workflow integration;
- Insightful tracking: Automatically publish results to PhariaStudio leaderboards for visualization and monitoring.
Requires Python 3.12+ and access to Pharia AI or OpenAI tokens. The supported models include GPT-4.1, Llama 3.3 70B Instruct, Llama 3.1 8B Instruct. Some advanced analytics and custom metrics are planned in future iterations.
Improved Portal Experience: Centralized access to all Applications
The updated Apps page in PhariaAssistant gives users a centralized, scalable view of all available Applications — including both default and custom-built tools — making it easier to discover and access the right capabilities quickly.
Key Benefits
- Improved discoverability: Easier navigation of all tools — no more hidden custom apps;
- Scales with your needs: Structured layout supports growing app libraries without clutter;
- Boosts adoption: Helps teams find and use the right apps faster, increasing tool engagement.
Compared to the previous unstructured grid view, this release introduces a unified, organized interface for managing a larger number of apps. It enhances usability for knowledge workers while laying groundwork for future improvements like role-based visibility and personalized layouts.
All users currently see the same app view. Features like favorites, personalized sorting, and access control are planned for Q4.
PhariaOS
In this release, we’re enabling flexible, standalone use across enterprise systems for the Translation capability. This broadens how PhariaAI features can be embedded into real-world operations.
Translation via Inference API: Easily Integrate Translation Capabilities into Your Own Workflows
The Translation worker capability is now accessible through a dedicated Inference API endpoint. This replaces the previously internal-only pharia-translate, enabling seamless and authenticated integration of translation functionality into custom enterprise applications and services — both within and beyond PhariaAI.
Key Benefits
- Seamless integration into enterprise workflows via a public, authenticated API;
- Improved extensibility by decoupling translation from the PhariaAssistant interface.
This update provides API-based access to the same underlying translation model used in PhariaAI, allowing customers to use translation independently in their own software environments.
No new translation features are introduced; existing limitations remain.
Other updates
- Bug fixes & improvements
- Security enhancements
- Minor performance and stability fixes