This changelog references changes made both to the Django backend, django-ai-assistant, and the
frontend TypeScript client, django-ai-assistant-client.
!!! note The backend and the frontend are versioned together, that is, they have the same version number. When you update the backend, you should also update the frontend to the same version.
- Add support for optional providers when installing the package
- Added compatibility with single-module Django apps
- Enabled langgraph astream usage
- Migrated coverage reporting to Codecov
- add Python 3.13 support across CI, tox, and project metadata
- Bump django-ninja, langchain, langgraph, langchain-openai and langchain-anthropic versions
- Fix save_django_messages on DBs where can_return_rows_from_bulk_insert=False
- INVALID due to publishing issues
- Improve support for models that do not accept the
temperatureparameter in theAIAssistantclass.
- Fix an
AttributeErrorraised in RAG AIAssistants when theretrieversupports theinvokecall only with the query string asinput.
- Refactor the code to use LangGraph instead of LangChain LCEL
(except for RAG functionality, see the
get_history_aware_retrievermethod). - Store all messages in the
Threadmodel, including tool calls and their outputs. - Allow separation of threads per assistant:
assistant_idin theThreadmodel. - New
updateThreadfunction fromuseThreadListhook. - Improved examples:
- Add markdown rendering to HTMX example.
- Better Movie Recommendation example.
- Better Tour Guide example.
- Fix frontend README.
- Less restrictive Python version in pyproject.toml. Support future Python versions.
- Add support for Django 4.2 LTS
- Add support for Python 3.10 and 3.11
- Initial release