Thanks to visit codestin.com
Credit goes to Github.com

Skip to content

v2.2.0

Latest

Choose a tag to compare

@schmitech schmitech released this 13 Dec 01:59
· 146 commits to main since this release

[2.2.0] - 2025-12-12

Core System Updates

  • vLLM Dual-Mode Support: Added direct mode to vLLM inference service allowing in-process model loading with GPU, similar to llama_cpp and bitnet providers
  • Audio Config Split: Split sound.yaml into separate tts.yaml and stt.yaml with independent provider configs, updated adapters to use stt_provider/tts_provider
  • Ollama Improvements: Added configurable adapter_preload_timeout (default 120s), improved warm-up with keep_alive and skip-if-loaded optimization, added ollama_remote for self-hosted servers
  • Language Detection: Skip Redis calls when stickiness is disabled, fixed confidence calculation and metadata exposure bugs
  • Jinja2 Templates: Converted intent templates to Jinja2 format with corresponding unit tests (Issue #69)
  • Qdrant Cloud: Added URL-based connection support with auto-index creation for payload fields
  • File Processing: Added MarkItDown as alternative processor, added full data mode for CSV/JSON with configurable thresholds for exact lookups
  • Adapter Config Fix: Fixed adapter config not updating after reload in chat service (Issue #92)
  • Maintenance Mode: Added runtime flag and UI for out-of-service messaging

Chat-app & UI Improvements

  • Mobile UI Enhancements: Added iOS/Android PWA meta tags, improved touch interactions, redesigned MessageInput with stacked layout, added larger touch targets (44px)
  • Dark Mode Fixes: Fixed desktop chat dark background inconsistencies, aligned chat layout with sidebar
  • Adapter Notes: Sync markdown styling with theme, fix notes not loading on startup in middleware proxy mode
  • Middleware Proxy: Auto-select first adapter on startup, replace adapters.yaml with environment variable configuration
  • Thread Improvements: Fixed markdown rendering in conversation threads

Bug Fixes & Technical Improvements

  • Middleware Proxy Security: Removed API keys and URLs from /api/adapters response, fixed undefined variable errors
  • Startup Fixes: Resolved startup errors, reduced duplicate logging, dynamically load providers from inference.yaml
  • CLI Fixes: Fixed arguments not overriding build-time env vars in orbitchat, fixed MaxListenersExceededWarning
  • Template Rendering: Fixed undefined variables in Jinja2 filters, fixed HTTP Retriever and GraphQL templates after migration
  • MongoDB: Fixed insertion error for documents containing ObjectApiResponse, deduplicate query results
  • File Cleanup: Fixed orphan files remaining after clearing conversations in middleware mode
  • Script Fixes: Fixed bash compatibility issues on Mac

API & Client Updates

  • orbitchat v2.3.9: Published new NPM package versions (v2.2.5 through v2.3.9) with UI improvements and bug fixes
  • API Middleware: Fixed integration for threads, files, and new conversations

Documentation & Configuration

  • AWS Deployment: Added ALB deployment guide with WAF configuration, Dockerfile and docker-compose for containerized deployment
  • Dependencies: Updated Ollama to 0.6.1, moved Ollama to default profile, updated tarball script with pre-configured db
  • Examples: Added Alberta Shelter Occupancy Adapter, replaced Contact with HR template examples
  • Installation: Reordered options with latest release as recommended method