Ogbujipt

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

Changed

- Deprecated `first_choice_text` & `first_choice_message` methods in favor of `first_choice_text` attributes on response objects
- Clarify set quite setup docs

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

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

Changed

- Use PG timestamp rather than serial for chat logs

Removed

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

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

Fixed

- `embedding_helper.py` logic