v0.48.0

PromptingTools v0.48.0

Diff since v0.47.0

Added

• Implements the new OpenAI structured output mode for aiextract (just provide kwarg strict=true). Reference blog post.

Commits

Merged pull requests:

• Add OpenAI structured outputs (#194) (@svilupp)

Closed issues:

• Broken on Julia 1.11beta2 (#158)
• Add support for OpenAI's Structured Outputs (#191)

v0.47.0

PromptingTools v0.47.0

Diff since v0.46.0

Merged pull requests:

• Add DTM specialized methods (#193) (@svilupp)

Closed issues:

• Unable to get started (#168)

v0.46.0

PromptingTools v0.46.0

Diff since v0.45.0

Merged pull requests:

• Update docstrings for RAGTools (#188) (@svilupp)
• New GPT4o (#190) (@svilupp)

v0.45.0

PromptingTools v0.45.0

Diff since v0.44.0

Breaking Change

• getindex(::MultiIndex, ::MultiCandidateChunks) now returns sorted chunks by default (sorted=true) to guarantee that potential context (=chunks) is sorted by descending similarity score across different sub-indices.

Updated

• Updated a hcat implementation in RAGTools.get_embeddings to reduce memory allocations for large embedding batches (c. 3x fewer allocations, see hcat_truncate).
• Updated length_longest_common_subsequence signature to work only for pairs of AbstractString to not fail silently when wrong arguments are provided.

Fixed

• Changed the default behavior of getindex(::MultiIndex, ::MultiCandidateChunks) to always return sorted chunks for consistency with other similar functions and correct retrieve behavior. This was accidentally changed in v0.40 and is now reverted to the original behavior.

Commits

Merged pull requests:

• Updates embedding concatenation (#186) (@svilupp)
• Fix getindex (#187) (@svilupp)

v0.44.0

PromptingTools v0.44.0

Diff since v0.43.0

Added

• Added Mistral Large 2 and Mistral-Nemo to the model registry (alias mistral-nemo).

Fixed

• Fixed a bug where wrap_string would not correctly split very long Unicode words.

Commits

Merged pull requests:

• Add mistral large 2 (#184) (@svilupp)
• Fix wrap_string for long Unicode (#185) (@svilupp)

v0.43.0

PromptingTools v0.43.0

Diff since v0.42.0

Added

• Added Llama 3.1 registry records for Fireworks.ai (alias fllama3, fllama370, fllama3405 and fls, flm, fll for small/medium/large similar to the other providers).

Commits

Merged pull requests:

• Add Fireworks.ai llama3.1 (#183) (@svilupp)

v0.42.0

PromptingTools v0.42.0

Diff since v0.41.0

Added

• Registered new Meta Llama 3.1 models hosted on GroqCloud and Together.ai (eg, Groq-hosted gllama370 has been updated to point to the latest available model and 405b model now has alias gllama3405). Because that's quite clunky, I've added abbreviations based on sizes small/medium/large (that is 8b, 70b, 405b) under gls/glm/gll for Llama 3.1 hosted on GroqCloud (similarly, we now have tls/tlm/tll for Llama3.1 on Together.ai).
• Generic model aliases for Groq and Together.ai for Llama3 models have been updated to point to the latest available models (Llama 3.1).
• Added Gemma2 9b model hosted on GroqCloud to the model registry (alias ggemma9).

Updated

• Minor optimizations to SubDocumentTermMatrix to reduce memory allocations and improve performance.

Commits

Merged pull requests:

• Register Llama 3.1 models + minor retrieval improvement (#182) (@svilupp)

v0.41.0

PromptingTools v0.41.0

Diff since v0.40.0

Added

• Introduced a "view" of DocumentTermMatrix (=SubDocumentTermMatrix) to allow views of Keyword-based indices (ChunkKeywordsIndex). It's not a pure view (TF matrix is materialized to prevent performance degradation).

Fixed

• Fixed a bug in find_closest(finder::BM25Similarity, ...) where the view of DocumentTermMatrix (ie, view(DocumentTermMatrix(...), ...)) was undefined.
• Fixed a bug where a view of a view of a ChunkIndex wouldn't intersect the positions (it was returning only the latest requested positions).

Commits

Merged pull requests:

• Add SubDocumentTermMatrix (#181) (@svilupp)

v0.40.0

PromptingTools v0.40.0

Diff since v0.39.0

Added

• Introduces RAGTools.SubChunkIndex to allow projecting views of various indices. Useful for pre-filtering your data (faster and more precise retrieval). See ?RT.SubChunkIndex for more information and how to use it.

Updated

• CandidateChunks and MultiCandidateChunks intersection methods updated to be an order of magnitude faster (useful for large sets like tag filters).

Fixed

• Fixed a bug in find_closest(finder::BM25Similarity, ...) where minimum_similarity kwarg was not implemented.

Commits

Merged pull requests:

• Add SubChunkIndex (view of index) (#179) (@svilupp)

v0.39.0

PromptingTools v0.39.0

Diff since v0.38.0

Breaking Changes

• Changed the default model for ai* chat functions (PT.MODEL_CHAT) from gpt3t to gpt4om (GPT-4o-mini). See the LLM-Leaderboard results and the release blog post.

Added

• Added the new GPT-4o-mini to the model registry (alias gpt4om). It's the smallest and fastest model based on GPT4 that is cheaper than GPT3.5Turbo.

Commits

Merged pull requests:

• Add GPT-4o-mini + set as default (#180) (@svilupp)