v0.4.61

• Bug fix - use full path rather than file name when creating a mapping between document name and its hash. Using file names only prevented files with identical names but different folders from being updated properly (e.g. index.html).
• WARNING - this fix requires re-indexing to update the hashes

v0.4.6

Feature - an ability to update embedding from the webapp (thanks @Hisma for suggestions and testing)

v0.4.5

• Feature - an ability to update embedding with changed (or new) documents, instead of reindexing from scratch. Using this feature requires reindexing the documents once using the new version.

v0.4.3

• Fix a bug in the webapp - when trying to switch to a new config, the old config is reloaded.

v0.4.2

• Implement feature #58 - an ability to switch between the configs in the web app (thanks @Hisma for suggestions and testing)

v0.4.1

• Feature - chat history in the web app (thanks @Hisma for suggestions and help with testing)
• Documentation update

v0.4.0

• Implements HyDE (Hypothetical Document Embeddings) - see https://arxiv.org/pdf/2212.10496.pdf
• To enable HyDE, use the following configuration block in semantic_search section.

hyde:
    enabled: True

v0.3.3

Add explicit CUDA options to fix GPU support for llama-cpp models

v0.3.2

• Fix bug that prevented finding documents using Python 3.11 based virtual environments.

v0.3.1

• Add progress indicators for embeddings generation
• Add support for Azure OpenAI
• Swith Google Colab notebook demo to use GGUF instead of GGML (which is retired)