CHANGELOG.md

Changelog

Notable changes to Format based on Keep a Changelog. Project follows Semantic Versioning.

[0.9.3] - 20240722

Added

• demo/chat_doc_folder.py - "chat my folder of docs" demo
• llm_wrapper.response_type to indicate tool-call LLM responses

Fixed

• llm_wrapper.llm_response object generation

[0.9.2] - 20240625

Added

• joiner param to text_helper.text_split() for better control of regex separator handling
• query filter mix-in, embedding.pgvector.match_oneof(), for use with meta_filter argument to DB.search
• llm_wrapper.response_type to differentiate tool calling vs regular LLM responses

Changed

• Index word loom items by their literal default language text, as well
• Cleaned up PGVector query-building logic

Fixed

• llm_wrapper.llm_response objects to handle tool calls
• failure of some matching scenarios in embedding.pgvector.match_exact()

Removed

• Previously deprecated first_choice_text & first_choice_message methods

[0.9.1] - 20240604

Fixed

• Demos & testing around text_helper.text_split_fuzzy()

[0.9.0] - 20240604

Added

• text_helper.text_split_fuzzy() as basically a generator version of text_helper.text_splitter(), and deprecate the latter
• text_helper.text_split() which brooks no overlap
• embedding.pgvector_data module for vector database embedding of data fields using PGVector. Document fields are now treated as simple text, with specialized metadata generalized in the new metadata field
• query filter mix-in, embedding.pgvector.match_exact(), for filtering vector search by metadata fields via new meta_filter argument to DB.search. This restores some of the tag matching search functionality that's been removed.
• clone() method on wordloom.language_item class (formerly text_item)
• meta and preserve_key args on wordloom.language_item (formerly text_item) initializer, to preserve TOML table items and top-level TOML key for each language item, respectively, in object properties

Changed

• Deprecation of text_helper.text_splitter()
• tags field of table encapsulated in embedding.pgvector_data now modified into a general, JSON meta field
• Word Loom now uses _ as the TOML table key for text content. The former text is deprecated
• Word Loom now uses _m as the TOML table key for text substitution markers. The former markers is deprecated
• Word Loom now reserves TOML table keys beginning with _, and passes on all other keys to the new meta property
• wordloom.text_item class now renamed wordloom.language_item
• lang arg on wordloom.language_item (formerly text_item) initializer renamed deflang

Removed

• embedding.pgvector_data_doc (now just embedding.pgvector_data)
• conjunctive option for tags searching now removed, in favor of a query filter mix-in approach

[0.8.0] - 20240325

Added

• llm_wrapper.llama_cpp_http_chat & llm_wrapper.llama_cpp_http; llama.cpp low-level HTTP API support
• llm_wrapper.llama_response class with flexible handling across API specs
• window init param for for embedding.pgvector.MessageDB, to limit message storage per history key
• threshold param on MessageDB.search()

Changed

• Deprecated first_choice_text & first_choice_message methods in favor of first_choice_text attributes on response objects
• Clarify test suite setup docs

[0.7.1] - 20240229

Added

• MessageDB.get_messages() options: since (for retrieving messages aftter a timestamp) and limit (for limiting the number of messages returned, selecting the most recent)

Changed

• PGVector users now manage their own connection pool by default
• Better modularization of embeddings test cases; using conftest.py more
• pgvector_message.py PG table timstamp column no longer a primary key

Fixed

• Backward threshold check for embedding.pgvector_data_doc.DataDB

[0.7.0] - 20240110

Added

• Command line options for demo/chat_web_selects.py
• Helper for folks installing on Apple Silicon: constraints-apple-silicon.txt
• Function calling demo
• embedding.pgvector_message.insert_many()

Changed

• Improved use of PGVector helper SQL query parameters
• PGVector helper search(query_tags=[..]) now uses contains operator (filters by existence in tag sets), not the same as where tags are OR
• PGVector helper search can now be set to work conjunctively or disjunctively
• PGVector helper query now has threshold arg based on degree of similarity. limit default now unlimited. Use SQL query args for query_embedding.
• embedding.pgvector split into a couple of modules.
• Separated data-style PGVector DBs from doc-style. tags is no longer the final param for PGVector docs helper methods & some params renamed.
• PGVector helper method results now as attr_dict
• PGVector helper now uses connection pooling & is more multiprocess safe
• embedding.pgvector_chat renamed to embedding.pgvector_message
• DB MIGRATION REQUIRED - embedding.pgvector_message table schema

Fixed

• insert_many PGVector helper method; semantics & performance
• demo/chat_web_selects.py & demo/chat_pdf_streamlit_ui.py (formerly non-functional)
• Tests & CI for PGVector helper

[0.6.1] - 20231114

Changed

• Use PG timestamp rather than serial for chat logs

Removed

• test/test_text_w_apostrophe.ipynb (incorporated into test/embedding/test_pgvector.py)

[0.6.0] - 20231113

Added

• Support for efficient multi-queries (executemany): insert_many vs insert
• Chatlog-specific PGVector helper (PGvectorHelper specialized into DocDB & MessageDB)
• PG Vector DB instance launch fo ruse in test suite & GitHub actions
• Updated model styles and prompt formatting, particularly for improved closed-context patterns & per-context metadata (e.g. for Airboros)

Changed

• Model introspection moved to llm_wrapper classes: hosted_model & available_models
• Move OAI API response structure handling helpers to be static methods of the llm_wrapper classes
• Clarified demo names
• Support upstream python-openai > 1.0

Fixed

• README sample code
• Demos
• Test cases
• Use of string formatting intead of SQL query parameters
• Registration of vector type
• pgvector test case

[0.5.1] - 20231010

Fixed

• embedding_helper.py logic

[0.5.0] - 20230919

Added

• Support for GGUF in download-model.py
• Support for in-memory LLM loading via ctransformers
• PostgreSGL vector support to embedding_helper.py, new class PGvectorConnection
  • PGvectorConnection is a wrapper around asyncpg, and is primarily just capable of excecuting raw SQL queries right now.
  • There are a few common SQL queries included in the class for using PGv as a vector database, but they are not yet fully tested.
  • Added a demonstration notebook which uses PGvectorConnection to do similarity search
• oapi_first_choice_content function

Changed

• Switch to a class-based wrapper for LLM endpoints/handlers - #39

Fixed

• Model style tweaks

[0.4.0] - 20230728

Added

• Initial implementation of Word Loom (see, for example, demo/multiprocess.py)
• More tests to improve coverage
• Qadrant collection reset method (embedding_helper.py)

Fixed

• Input parameter controls in embedding_helper.py

Changed

• test suite structure

[0.3.0] - 20230723

Added

• __version__
• chat_web_selects.py demo
• async_helper.save_openai_api_params()

Fixed

• chat_pdf_streamlit_ui.py demo
• OpenAI API reentrancy & async_helper.py

Changed

• Renamed demo alpaca_simple_fix_xml.py → simple_fix_xml.py
• Renamed demo alpaca_multitask_fix_xml.py → multiprocess.py
• Renamed oapi_choice1_text() → oapi_first_choice_text()
• Renamed async_helper.schedule_llm_call() → async_helper.schedule_callable()

[0.1.1] - 20230711

Added

• GitHub CI workflow
• Orca model style
• Convenience function oapi_choice1_text()
• Additional conveniences in prompting.model_style

Fixed

• Linter cleanup

Changed

• Qdrant embeddings interface
• Renamed prompting.context_build() → prompting.format()

[0.1.0]

• Initial standalone release candidate