NEWS

---

BFHllm 0.2.0

Nye features

• Skiftet default LLM-model fra gemini-3.1-flash-lite-preview til stable gemini-3.1-flash-lite. Sync af model-ID på tværs af docstrings, README, eksempler og tests (tidligere drift mod gemini-2.5-flash-lite rettet).
• Tilføjet GitHub Actions workflows der spejler biSPCharts' versioneringspolitik:
  • R-CMD-check.yaml — smoke check på alle pushes/PRs til main
  • tag-release.yaml — automatisk vMAJOR.MINOR.PATCH-tagging efter grøn R-CMD-check på main. MAJOR.MINOR læses fra DESCRIPTION (manuelt bump i PR), PATCH auto-inkrementeres pr. merge.

Bug fixes

• Rettet stale test-assertions for default-config (gemini-2.5-flash-lite → gemini-3.1-flash-lite, timeout_seconds = 10 → 120) som har været brudt siden config-bumps i v0.1.x.

BFHllm 0.1.2

Bug fixes

• Tilføjet Remotes: for BFHcharts (Suggests). Downstream-installationer kunne tidligere fejle ved --as-cran tests der forsøger at installere Suggests. Fra v0.1.2 er transitiv dep-resolution konsistent med sibling- pakkerne (BFHcharts v0.8.1).

BFHllm 0.1.1

Bug fixes

• Fixed deprecation warning by updating to credentials parameter in ellmer::chat_google_gemini() and ragnar::embed_google_gemini() calls (was api_key prior to ellmer 0.4.0)
• Fixed roxygen2 deprecation warning by removing @docType package and documenting "_PACKAGE" instead
• Updated ellmer dependency to >= 0.4.0 to ensure compatibility

BFHllm 0.1.0

Initial release

• Core LLM chat interface with Google Gemini provider
• Provider abstraction layer for future model support
• Configuration management system with defaults
• Circuit breaker pattern for API resilience
• Response validation and sanitization
• Generic caching layer with Shiny session adapter
• RAG integration with ragnar knowledge stores
• SPC-specific improvement suggestions module
• Comprehensive test suite with 142 passing tests
• Examples for basic chat, SPC suggestions, and Shiny integration