django-ai-assistant
django-ai-assistant copied to clipboard
Integrate AI Assistants with Django to build intelligent applications
(some dependencies seem to still be broken)
To avoid errors due to token limit. See: - https://langchain-ai.github.io/langgraph/how-tos/memory/manage-conversation-history/ - https://python.langchain.com/docs/how_to/trim_messages/ - https://langchain-ai.github.io/langgraph/concepts/memory/?ref=blog.langchain.dev#summarizing-past-conversations
https://langchain-ai.github.io/langgraph/how-tos/node-retries/ Check also retries on `ToolNode` and chat model instantiation.
- Talk about the architecture: we first run tools to gather information and then call the model requiring the structured output. - Mention that using pydantic is better than typeddict...
From the [quick start](https://langchain-ai.github.io/langgraph/tutorials/introduction/#part-1-build-a-basic-chatbot) ``` from IPython.display import Image, display try: display(Image(graph.get_graph().draw_mermaid_png())) except Exception: # This requires some extra dependencies and is optional pass ```
This was left out during [migration of the core flow](https://github.com/vintasoftware/django-ai-assistant/pull/165) but we should have everything fully ported to the new patter.
Must have a way to let users bring their own API key, such as https://gptlab.streamlit.app/ Host a frontend-only app on GitHub pages?
Useful for removing TourGuideAssistantView and having the frontend to call the assistant directly.
https://www.anthropic.com/news/prompt-caching Related to: https://github.com/vintasoftware/django-ai-assistant/issues/160
Ideas: - Complex AND/OR filtering via Django admin that generates a special query * https://github.com/cloudblue/django-rql * https://github.com/ivelum/djangoql - And exports a CSV or Excel - Some application that generates json...