PhariaAI v1.250800.1

Release Version

v1.250800.1

Release Date

August 20, 2025

Availability

On-premise and hosted


PhariaAI stack updates

Multimodal Support in PhariaInference API: Prompts with Text & Images

Availability

On-premise and hosted

The PhariaInference API now supports multimodal prompts, allowing developers to send both text and images in requests to the /complete and /chat endpoints. This unlocks richer interactions, enabling use cases where visual context is essential. The /chat endpoint remains fully OpenAI-compatible.

Key Benefits:

  • Combine text and images in prompts for more contextual, multimodal responses.

For further information, see the documentation.

The images must be base64-encoded before sending. External image URLs are not yet supported.

Bug fixes

For PhariaOS:

  • Fixed issue causing undeployment requests to timeout.

  • Fixed issue causing 401 errors on application load.

For PhariaAssistant:

  • Fixed an error that occurred when uploading large multipage PDFs in Chat.

  • Enabled Build command support on Windows in PhariaCLI.

  • Fixed an issue in Safety skill that caused a 500 server error when generating summaries.

  • Fixed an issue where summaries were returned as JSON instead of properly formatted Markdown.

Security enhancements

For PhariaOS:

  • Resolved CVE-2025-47907.

Other updates

  • Updated to Bitnami legacy images for third-party dependencies (Postgres, RabbitMQ, Redis, MinIO). See linked issue for details.

  • Made base_repository optional when configuring a PhariaEngine namespace.

  • Minor performance, UX, and stability fixes.