Skip to content

Releases: timescale/pgai

pgai: v0.7.0

06 Feb 08:33
Compare
Choose a tag to compare

0.7.0 (2025-02-05)

Features

  • add vectorizer enable/disable support for ai.scheduling_none (f3d91a3)

Bug Fixes

  • remove sqlalchemy warning about conflicts (#413) (55f89fe)

extension-0.8.0

04 Feb 19:03
Compare
Choose a tag to compare

New features and improvements

  • Add vectorizer enable/disable support for ai.scheduling_none (#402)
  • Add LiteLLM embedding and vectorizer integration (#320)
  • Add Ollama support for tool use and structured outputs (#403)
  • [BREAKING] Update cohere embed, chat, rerank to use V2 endpoints (#417)
  • Add ai.anthropic_list_models (#386)

Fixes

  • Changed tool_choice type and null args omitted (#405)

Other improvements

  • Update anthropic library to 0.44.0 (#385)
  • Improve and split DEVELOPMENT.md (#380)

pgai: v0.6.0

28 Jan 09:26
9a5424a
Compare
Choose a tag to compare

0.6.0 (2025-01-28)

Features

  • disable OpenAI tokenization when a model does not have a tokenizer match (#390) (41cb52c)
  • add LiteLLM vectorizer integration. Requires a compatible version of the PGAI extension, which is currently under development (#320) (0fb7e46)

Miscellaneous

  • get rid of nested parametrization (#394) (0a399e2)
  • separate test_vectorizer_cli.py into separate files for vectorizer types (#401) (c64833c)

pgai: v0.5.0

23 Jan 13:21
fb02a19
Compare
Choose a tag to compare

0.5.0 (2025-01-22)

Features

  • add alembic operations for vectorizer (#266) (b01acfe)
  • allow users to configure a base_url for the vectorizer OpenAI embedder (#351) (66ceb3d)

Bug Fixes

  • two usability issues with sqlalchemy (#354) (95fa797)
  • vectorizer_relationship for sqlalchemy models with mixins or inheritance (#357) (cfd5f73)

extension-0.7.0

20 Jan 09:59
31f2e34
Compare
Choose a tag to compare

New features and improvements

  • Allow users to configure a base_url for the vectorizer OpenAI embedder (66ceb3d)
  • Upgrade ollama client to 0.4.5 (c579238)
  • Add just ext docker-start command (96ac4f5)
  • Allow vectorizers to be granted to public (7b2995b)
  • Allow superusers to create vectorizers on any table (027b3f4)

Fixes

  • Fix load_datasets to handle structs (25465ae)
  • Fix deprecation warning on re.split (7b4a916)
  • Fix exclude python system packages for versioned extension (1f6d1a8)
  • Fix schema qualify type definitions, casts, and operators (ee86d35)
  • Fix host networking not supported on macOS (aac3d83)
  • Fix handling of empty PG_BIN (bd83165)

Other improvements

  • Add warning when trying to install pre-release version (0b400a0)
  • Support uv in extension install for development (3f9736a)

pgai: v0.4.0

10 Jan 14:38
Compare
Choose a tag to compare

0.4.0 (2025-01-08)

Features

  • add sqlalchemy vectorizer_relationship (#265) (0230509)
  • load api keys from db in self hosted vectorizer (#311) (b7573b7)
  • print unexpected error traceback in debug logs (#344) (d9bdcd6)
  • pull missing ollama models (#301) (dbac246)
  • upgrade ollama client to 0.4.5 (#345) (c579238)

Bug Fixes

  • handle 'null' value in chunking 'chunk_column' (#340) (f283b6c)

Miscellaneous

  • fix broken pgai build by pinning hatchling (#308) (5441f2d)
  • register postgres_params custom pytest.mark (#327) (89039b2)
  • split embedders in individual files (#315) (77673ee)

pgai: v0.3.0

10 Dec 14:43
0146781
Compare
Choose a tag to compare

0.3.0 (2024-12-10)

⚠ BREAKING CHANGES

  • remove truncate parameter from Ollama/Voyage APIs (#284)

Features

  • add Voyage AI vectorizer integration (#256) (1b56d62)
  • remove truncate parameter from Ollama/Voyage APIs (#284) (ecda03c)
  • add load_dataset function for creating a table from a HuggingFace dataset (#253)(2946938)

Bug Fixes

  • fail fast when api key is missing and once is set (#274) (1c2ff20)

extension-0.6.0

17 Dec 11:33
77f3863
Compare
Choose a tag to compare

This release adds support for using Voyage AI in a vectorizer, and loading
datasets from hugging face.

New features and improvements

  • Use the ai.voyageai_embed, and ai.embedding_voyageai functions to use Voyage AI for vector embeddings 1b56d62.
  • Add ai.load_dataset to load datasets from hugging face 2946938.
  • Change the type of keep_alive parameter from float8 to text 0c74741
  • Remove truncate parameter from Ollama/Voyage APIs ecda03c

Fixes

Full Changelog: extension-0.5.0...extension-0.6.0

pgai: v0.2.1

02 Dec 11:31
Compare
Choose a tag to compare

0.2.1 (2024-12-02)

Bug Fixes

pgai: v0.2.0

26 Nov 16:25
Compare
Choose a tag to compare

0.2.0 (2024-11-26)

Features

  • add Ollama support to vectorizer (6a4a449)

Bug Fixes

  • make vectorizer worker poll for new vectorizers (0672e7a)

Miscellaneous

  • a new database for each test (4ed938b)
  • add logo to pgai pypi (3366368)
  • add test for recursive text splitting (#207) (4a35fc6)
  • migrate project commands from Make to Just (42a8f79)
  • migrate to uv and hatch (#188) (627cf33)
  • refactor test infra (ac845ca)
  • run pgai tests against extension from source (ffc20d2)
  • scope postgres_container fixture to class (12c1780)
  • separate the dev/test/build between the projects (183be9e)
  • test the cli instead of the lambda handler (#204) (3a48f82)